The Alpha School Debate
The network of private schools may have harnessed the power of AI to dramatically boost test scores, but is that all there is to education?
Alpha School—a chain of private schools that claims to have revolutionized education with the help of AI—has been around since 2014, but it’s having a moment. With campuses rapidly sprouting in affluent parts of the country, it’s claiming to enable students to learn at a rate that is 2.6 times faster than average while spending only two hours a day on academics.
During those two hours, each student is served a personalized menu of lessons. Rather than assuming that just because a student is in, say, fifth grade, they should be doing fifth-grade work, Alpha uses an AI-powered app to pinpoint where the student needs to begin—either higher or lower than “grade level”—and provides appropriate material. The app then provides prompt, targeted feedback. When students have mastered a particular level of difficulty, it gives them more challenging material.
There are no human teachers. Instead, Alpha provides “guides,” whose role is to motivate students to do their best with the app.
Alpha says this approach results in dramatically faster growth for students at any ability level, as measured by standardized test scores. For elementary and middle grades, those tests are MAP assessments—an acronym for Measures of Academic Progress—that are given to millions of students across the country throughout the school year. Most Alpha students are in the top one percent for growth on the MAP. At the high school level, outcomes are measured by the SAT and AP tests. The average SAT score at Alpha is 1470, and 90 percent of students get 4s or 5s on the AP.
Academics occupy the morning at Alpha School. What about all that “time back,” as Alpha calls it, that students get in the afternoon—the additional four hours or so that, in a traditional school, would also be spent on academics? That’s when kids at Alpha get to participate in group “workshops” designed to teach “life skills.” They can choose to learn how to assemble furniture from IKEA, put together a Rubik’s cube, engage in juggling or public speaking—or even, in at least one instance, own and operate an Airbnb.
Division Within the Science of Learning Community
Alpha School and its parent company, 2-Hour Learning, argue that the tired, inefficient era of “a teacher standing in front of a classroom” must come to an end. Perhaps not surprisingly, not everyone agrees—and I have some serious reservations myself.
Before I summarize the critiques of others and add my own, I should mention that I have never visited an Alpha School, nor have I been able to get any official information from 2-Hour Learning or anyone at Alpha. I submitted an inquiry but have not heard back, and I’ve been advised by another writer who has been trying to reach them that I should not hold my breath. Still, I’ve reviewed lots of publicly available material, including a lot from Alpha itself. If I should hear anything from 2-Hour Learning or Alpha that is relevant to this post after it’s published, I will update or correct it.
Alpha has claimed they have “globally renowned learning scientists” helping them build their model, based on decades of research. And indeed, at least two prominent advocates for science-aligned instruction, Carl Hendrick and Zach Groshell, are consulting with Alpha and helping them develop their instructional apps. Hendrick says he started out a skeptic but came away from a visit convinced that Alpha was engineering “the best conditions for learning through curriculum design and learning science.” Groshell has called Alpha’s approach “cognitive load theory in a bottle.”
Others within the science-of-learning movement remain skeptical, however, for various reasons. Jared Cooney Horvath is doubtful that success with an AI-assisted tutor, or on tests that mimic that situation, will transfer to other learning contexts. Robert Pondiscio questions Alpha’s confident, even arrogant tone and points out they haven’t made their underlying data available for review by independent third parties. Dylan Kane notes that learning via a screen is fundamentally different from human-powered instruction; for example, studies have shown that reading on screens results in weaker comprehension.
Paul Kirschner, who has co-authored two books with Hendrick, agrees with all of the above and adds that academic subjects shouldn’t be viewed as “just administrative hurdles that kids have to clear before their ‘real’ education and learning can begin.” In addition, several people have questioned whether Alpha’s model would work with students from less affluent and highly educated families. Tuition at Alpha can run as high as $75,000 a year.1
Before I get to my own doubts, I’ll acknowledge that the Alpha model does appear to align with cognitive science in many respects. As Alpha co-founder Mackenzie Price has argued, an AI tutor can do things classroom teachers generally can’t—like meeting each student where they are and providing the specific level of challenge they need. Unlike human beings, AI is endlessly patient and non-judgmental, offering just-in-time explicit instruction and retrieval practice—techniques grounded in cognitive science principles that few teachers learn about during their training.
An Amazon Cart of Pink Stuff
And Alpha seems to have solved a fundamental problem that has plagued instruction-by-computer: lack of motivation. Generally speaking, a human teacher is more likely than a digital device to inspire students to apply themselves. One source of motivation at Alpha is the desire to get more “time back” by mastering the material at a rapid clip. Another is the guides, who act as cheerleaders. Yet another, which is less publicized: rewards for achievement.
In one online Alpha video, Price interviews an Alpha graduating senior named Elle who recalls that when she first started at the school in third grade, she was only in the 50th percentile on the science MAP test. Elle’s guide assured her that she could get to the 99th percentile and promised to buy her an Amazon cart full of pink merchandise she had picked out if she hit the 90th (she was into the color pink at the time). She reached the 95th percentile eight weeks later and claimed the cart.
In the video, Elle says that it wasn’t just the merchandise that motivated her but the confidence her guide expressed in her ability. Still, there was a lot of pink merchandise in the mix.
Evidence suggests that this kind of “extrinsic” motivation doesn’t last once the rewards disappear. And some argue that if people are already intrinsically motivated, extrinsic rewards can backfire. People may think: if this stuff is so unpleasant that I have to be bribed to do it, maybe I don’t actually want to do it without the bribe.
But one Anonymous Alpha Parent, who I’ll call the AAP, has written that even though his kids are self-motivated to learn, the incentive system has “turbocharged” their learning without reducing their internal drive. Of course, that’s far from a scientific study.2
My own critique of the Alpha model rests partly on some details and partly on fundamental principles. I’ll start with the details, some of which are significant.
What About the Humanities?
My sense is that AI-assisted tutoring works best for math, and maybe science, where learning proceeds in a more clearly hierarchical, sequential fashion. With math, as Price says in one video, “If you only get 70% of division problems correct you’re gonna really struggle in fractions, which require mastery of division and multiplication.” With literature or history, the sequence of learning isn’t as clear, and the universe of concepts to be learned is far less finite.3
Indeed, Alpha has called math its “crown jewel,” showing growth even for students who perform at the highest levels. But it’s not just that the Alpha model works best for math. It’s not clear to what extent Alpha even tries to teach anything other than math, science, reading—and writing, which I’ll get to shortly.4
When I left a comment on Paul Kirschner’s Substack raising this question, someone going by the moniker of “turing hamster,” who identified himself or herself as “an engineer on the literacy team at Alpha,” assured me that “all the main academic subjects are covered in the Alpha / Timeback curriculum: Reading, Language, Math, Writing, Science, Social Studies.” I asked for more detailed information but didn’t get a response.
Meanwhile, none of the Alpha videos I’ve watched say anything about social studies, with one exception: a video showing that within the two-hour academic block, students spend 25 minutes on social studies.5 Repeatedly, however, the eye-popping scores that Alpha boasts about are limited to math, reading, science, and sometimes writing. That may be because standardized tests like MAP don’t cover social studies content, which varies from state to state. In any event, we don’t really know how dramatically Alpha students are “growing” in terms of social studies knowledge, or if they’re growing at all.
How Is Social Studies Taught?
The AAP’s comments on social studies are somewhat contradictory. “Instead of classes in history, geography, economics and political science,” they write at one point, Alpha covers that material through passages in its AlphaRead reading comprehension tool. At another point, the AAP says that social studies—along with science and “language,” are “mostly on iXL.” A typical lesson, the AAP writes, consists of a YouTube video followed by problem sets with “third party tools.”
I Googled iXL and discovered it’s an online personalized learning tool—or, as the website claims, a “comprehensive K-12 curriculum.” For each grade, four subjects are listed: math, language arts, science, and social studies. Next to each subject is the number of “skills” covered.
For fifth grade, for example, social studies covers 123 “skills.” These include reading a map—which is, yes, a skill—but also content like “The Thirteen Colonies,” “World War II,” and “American symbols, landmarks, and monuments.” If you hover over one of the subtopics, you find a brief text relating to, say, “New England colonies: religion and government,” followed by multiple choice questions.
What About Whole Books?
I saw no indication in publicly available information that Alpha students read entire novels or any whole books—although one Alpha parent quoted in an article in The Cut insisted, “My kids do read books.”
Zach Groshell, who is helping Alpha write its math apps, has said that he wouldn’t send his own daughter to Alpha “because she loves reading whole books.” More recently, Groshell said on X that he is sending his daughter to an Alpha school, but it’s a specialized one that follows Montessori principles, Alpha’s “Montessorium.” He added that he’s not sending her there because of whole books, “which are across models,” but because of “fit.”
The head of the Montessorium, Matt Bateman, told me in an email that at his school, which serves children ages 5 and 12 years, “every student is assigned five or six whole works of literature per school year.” He added that “when a student is in a literature circle” they meet twice a week in groups of two to four for discussions led by a guide, along with written responses.
Other Alpha campuses, Bateman said, include whole-book reading on students’ “check charts,” which list certain tasks to be done outside of the two-hour academic block. “There’s an assessment for each check chart item,” he told me, “including books.”
Bateman said that he’s working on two “pilots” focused on reading whole books across the network in a way that aligns with Alpha’s model. But it’s not clear that model can accommodate a group of students reading and discussing a common novel because it is (rightly, Bateman believes) “skeptical about whole-class pacing.”
According to the AAP, some parents at the flagship Alpha school in Austin wanted more academic afternoon workshops—not just “life skills,” but subjects like writing, history, and economics. They were told they should try sending their kids to Alpha’s school for gifted and talented students.6
Again, Alpha’s model may work best for math and science, where the content lends itself more to right and wrong answers, than to the humanities. There are, of course, right and wrong answers about historical facts and the plot elements of novels. But there’s also a lot of nuance and questions that should be up for debate—not just with an AI tutor but with a teacher and classmates.
In the course of such discussions, students should be articulating and defending their ideas, as well as listening to arguments that go in a different direction and learning how to engage in civil disagreement. They should also have the opportunity to be collectively transported, through the medium of fiction, to another time and place—to share laughter and maybe tears with others. I’m not convinced a bot can provide that experience.
How Does Alpha Teach Writing?
For writing, Alpha uses its own proprietary program, AlphaWrite, to which outsiders are not allowed access. (I tried.) I did, however, manage to glimpse a screen showing a writing activity in an Alpha promotional video. To my surprise, the activity resembled one in a book I co-authored called The Writing Revolution.
On the screen was this directive: “Write a set of sentences using the conjunctions ‘because,’ ‘but,’ and ‘so.’”7 The sentence “stem” students were to work with was “Drums are an essential part of a band …” One of the suggested answers was, “Drums are an essential part of a band because they make a lot of noise.”
Leaving aside whether that answer makes sense, one problem is that the activity isn’t embedded in any curriculum content. It’s fine to introduce a new concept using everyday content like “drums,” but a key principle of The Writing Revolution method is that it should be embedded in the content of the curriculum as much as possible. Writing about content has the potential to deepen comprehension in a powerful way. Why waste that potential on random sentences about drums?
The AAP acknowledges that AlphaWrite is weak and the school is “taking it back to the drawing board.” As of the time he was writing, in June of 2025, he felt that Alpha students had no “real edge” over students in traditional schools as far as their writing skills were concerned.
Chatbots Are “Cheatbots”
Writing, as I’ve suggested, isn’t just a skill but also a potentially powerful lever for learning. And one risk of an AI-powered academic program is that students will use AI to subcontract out their writing, thereby missing out on a lot of learning. Mackenzie Price seems to be aware of this, earning kudos for saying that will never happen at Alpha. “Chatbots in education,” she has said, “are cheatbots.”
And yet: A reading specialist named Savannah Ngo has written about her visit to an Alpha School campus in California, where she was assigned to work with some children in K-2 during the “life skills” portion of the day. The kids were reading aloud books they had written, in preparation for doing so before an audience of parents, and Ngo was giving them feedback on their reading fluency and volume. So far, so good.
“But the snag came,” Ngo wrote, “when I sat down with the first student who immediately clarified to me that he had not written this book but actually he just typed in the computer and AI wrote it. As soon as he started reading, this became obvious. The story was saccharine and tidy in that artificial way that AI does so well and I’m afraid the child learned almost nothing from the experience. It might as well have been an older sister who wrote the book for him and told him to pass it off as his own.”
A one-off? Possibly. But even if Alpha students generally aren’t using AI to do their writing for them, it’s not clear what kind of writing they’re being asked to do and who, or what, is evaluating it. Do they write essays? Research papers? Does any human being read them?
Fundamental Problems with the Model
Even if Alpha figures out a better way to cover history, literature, and writing through its AI tutors, I would remain skeptical about the model. To be clear, I’m not against AI-assisted learning, period. The evidence indicates it can be highly effective in certain circumstances, including some of the ones in which Alpha uses it. I just don’t think we should hand over academic instruction entirely to bots.
Why limit human interaction to “life skills” workshops rather than integrating it into academic learning? Yes, using human teachers may be less efficient, but education isn’t solely a matter of efficiency. Some of its most important aspects, like learning to bounce around ideas about academic topics with one’s teacher and fellow students, may require something other than the shortest path to acquiring information.
Plus, as others have pointed out, Alpha’s model treats the “learning” part of school as something to be gotten through as quickly as possible in order to get to the good stuff. Is that really the message we want to send to kids about the life of the mind? Isn’t there also some good stuff to be found there?
You might say you can’t argue with success: I mean, look at those scores. In one video, Price boasts that graduates of Alpha have been admitted to colleges like Stanford, Vanderbilt, and USC. And the videos of Alpha students, including those at the high school level, show them to be bright, enthusiastic, articulate kids.
But test scores don’t measure everything. Yes, I and others have argued that standardized reading comprehension tests essentially reflect knowledge and vocabulary. But that doesn’t necessarily mean that students who ace multiple-choice questions on the SAT, or even the AP—which at least requires writing—are capable of mining a complex text for meaning, reflecting on it deeply, and expressing themselves coherently. Like others writing about education, I’ve spoken to quite a few college professors—including some at highly selective colleges—who have told me that in recent years their students are no longer capable of doing such things, despite presumably having high test scores.
It’s clear that many students and their families are happy with their experience at Alpha. But I have to wonder how graduates will fare in college, where—at least for now—they will encounter, perhaps for the first time, the supposedly antiquated model of “a teacher standing in front of a classroom.” How will they adjust to lectures, seminars, and more than three hours a day of academic work? Maybe some will do fine, but I suspect others will struggle.
Nevertheless, Price argues that Alpha’s model will work for almost all students and will spread along with growing interest in its apparent success. That may be happening already. Last year its parent company, 2-Hour Learning, got approval to open a tuition-free virtual charter school in Arizona.
But based on the information I have, I would advise caution. Even if Alpha’s data on growth in test scores is solid, those scores may not reflect the kind of education that can enable all students to reach their full potential.
In one video, Price acknowledges that sometimes the Alpha approach doesn’t work—including when parents aren’t “philosophically aligned with the model.” If I had school-age children, at this point I would put myself in that camp. And I suspect quite a few other parents would as well.
Update, 7.26.26: A clarification about Zach Groshell’s decision to send his daughter to Alpha’s Montessori school was added.
Update, 7.27.26: Additional comments from Matt Bateman about Alpha’s approach to whole books were added.
Alpha does have a school in Brownsville, Texas, where the poverty rate is about 24 percent, roughly twice the national average. They say students at that school learn 2.1 times faster than average. I haven’t seen data on whether those particular students are actually from low-SES families.
Alpha seems to have institutionalized the reward system since Elle was in third grade. Rather than carts of pink stuff from Amazon, the AAP’s kids earn “Bucks,” which they can spend at a school store.
On the other hand, math expert and cognitive science advocate Anna Stokke has argued, in a comment on a Substack post, that even math teaching requires “human intuition and human knowledge.”
In the video in which Elle talks about boosting her science score because of the promise of pink stuff, she and Price agree that science, in K-8, “is really just reading comprehension.” Price says a “fun fact” that has shaped Alpha’s philosophy is that science “is one of the easiest subjects to get to the 99th percentile.” For that reason, she says, “we will use science as a confidence anchor.”
The video suggests that, as in many traditional schools, social studies is given low priority: “Science and language take up about half the time that math and reading do,” Price says. “Once a student is finished with their science [and] language material, we give them history, social science and writing lessons.” So apparently, if a student needs to spend more than the allotted time on other subjects, they may not make it to history, social science, and writing.
The AAP also listed some of the workshops available at the gifted and talented school, including debate and chemistry for younger children and chess and robotics for older ones. History, writing, and economics weren’t mentioned.
On the screen, the closing quotation marks for each conjunction were incorrectly placed before the comma rather than after it. I don’t mean to quibble, but this is an activity that’s supposed to be teaching kids how to write—including how to correctly punctuate.
This is one of the clearest and most digestible critiques of Alpha that I’ve seen. I lost patience with the model—and its boosters—once I learned that a student could attend an Alpha school through 12th grade without ever being required to read a complete book.
I’m aware that students have choice time in which many of them read. But there is a reason choice-driven, minimally instructed literature circles do almost nothing for student growth. They are not an adequate substitute for a coherent, teacher-led literature curriculum.
I wrote about a related concern last November: even if we don’t believe Alpha’s model should become the spine of an entire school, large parts of its underlying philosophy have already spread through public schools in the form of programs like iREADY, IXL and Kahn Academy. They have been propelled by similarly dramatic claims of rapid growth, along with the ease and low cost of implementation for administrators.
https://www.middleschoolliteracyproject.org/p/ai-schools-are-selling-a-dangerous
"In addition, several people have questioned whether Alpha’s model would work with students from less affluent and highly educated families. Tuition at Alpha can run as high as $75,000 a year."
Appreciate all of the precision within this piece, but it feels like this should be a bigger point: shouldn't the comparison point be how Alpha compares to other private schools with similarly-lofty tuitions?
That feels like more than just a small detail to include, especially when extrapolating whether this "model" would work on a broader scale.