Alpha School Claims Learning Efficiency Doubled Faces Data Scrutiny
The American private school network Alpha School promotes an AI-driven model where students only need two hours of academic courses each day to achieve learning progress at twice the speed of regular schools. However, a review of its internal data by The New York Times shows that the actual percentage of students reaching this goal is significantly lower than advertised.
The network charges up to $75,000 per year, with students completing core subjects like math and reading through adaptive learning software for two to three hours in the morning, followed by workshops in entrepreneurship, public speaking, and sports in the afternoon. Traditional teachers are replaced by Guides who are responsible for motivation and coaching, and the school emphasizes no homework and minimal paper materials.
Data from tests conducted from Fall 2025 to early 2026, obtained by The New York Times, indicates that about 75% of students made progress at or above the national average. However, only 42% achieved double progress in both math and reading simultaneously, and about half of K-8 students did not meet the double progress target in math.
The performance of students with weaker foundations is even more limited: among those who previously scored at or below the national median in math, only 37% met the double progress standard. The overall median scores of students remain high, with math around the 87th percentile and reading around the 89th percentile.
Founded by MacKenzie Price and others, Alpha School has rapidly expanded in recent years, entering public education and Texas voucher programs, while gaining attention from the Trump administration and some business leaders. Its core platform, TimeBack, has served over 7,000 students across private, public, and home learning settings.
High tuition and marketing promises attract high-income families and policy attention, directing funds towards software platforms and expansion. However, the data gap raises questions among some parents and educators about the actual effectiveness; traditional schools and lower-cost options face comparative pressure, while families of students who do not meet promises bear opportunity costs.
Source: Public Information
ABAB AI Insight
Alpha School was founded in Austin in 2014, initially using adaptive software rather than generative large models, later strengthening its AI narrative and introducing the Guide role. Founders MacKenzie Price and Joe Liemandt, who has a software background, are involved, with the latter committing around $1 billion to support related AI products. The school has expanded from a few campuses to multiple cities and entered public and voucher systems.
In terms of capital, the school covers operations and software development with high tuition (most campuses charge $40,000 to $65,000, with San Francisco and Palo Alto reaching $75,000), while seeking public funds and voucher projects to scale. The motivation is to compress academics into two hours to free up time for life skills, and to attract tech and policy resources through AI narratives, pushing a transition from high-end private to broader coverage.
This is comparable to early adaptive learning platforms like IXL or Khan Academy's personalized paths, as well as some micro-schools and alternative education experiments. It is currently transitioning from marketing-driven expansion to accepting external data scrutiny, while traditional public and private schools face competitive pressure from efficiency narratives.
Essentially, this represents a limited realization of technological substitution: AI and adaptive software can accelerate mastery of some fundamentals and improve median performance, but it is difficult to universally deliver on the "double progress" promise, especially for weaker students; the mechanism involves selection effects and high-motivation environments elevating overall scores, while the true growth distribution is closer to conventional, leading to a systematic gap between advertising and internal data.
ABAB News · Cognitive Laws
Promised multiples do not equal actual distribution
High tuition buys narrative, not guarantee
Technology raises the baseline, fundamentals determine the slope