A look at Alpha School, a network of private schools that claims to teach kids 2x as fast using AI; data shows ~50% of K-8 students missed this goal in math
Alpha School, a private network, is spreading into public education. While its students' test scores are high, some parents and educators say it doesn't live up to the hype.
New York TimesDana Goldstein
Context & Ripple Effects
Alpha School had already been presented as part of a premium market in which high-income families use AI-powered private schools and tutors for tailored curricula and life-skills instruction, as described in a July look at AI-powered private education. Its move into public education therefore turns a private-school proposition into one that public systems can evaluate against stated learning targets.
The wider school-AI push has already produced a cautionary precedent: Newark Public Schools' Khanmigo trial raised concerns about chatbots taking over students' thinking. Alpha's internal math data gives parents and education decision-makers a more concrete basis for judging whether accelerated instruction matches its marketing.
First-order effects
Alpha School's claim that AI instruction teaches children twice as fast faces an immediate credibility test because about half of its K-8 students missed the network's math goal, even as students' overall test scores are high.
Parents and public-education counterparts considering Alpha's expansion gain outcome data that separates strong test performance from achievement of the network's faster-learning benchmark.
Second-order effects
AI education providers seeking school adoption will face pressure to define learning targets and disclose whether students meet them, rather than relying on broad claims of personalization or high scores.
The concerns raised in Newark's Khanmigo pilot become more salient for Alpha's public-school expansion: educators must assess both measured progress and how much instructional work the AI takes from students.
Third-order effects
If public-school adoption continues, AI tutoring is likely to be judged less by the presence of an AI tool than by accountable evidence that it improves learning against explicit goals.
The gap between Alpha's accelerated-learning claim and its disclosed math results points toward a more evidence-driven market for AI-enabled instruction, especially where public systems are involved.
The trend: AI education is moving from premium, personalized offerings toward public-system deployment, where product claims increasingly require transparent learning-outcome evidence.
My deep dive on the “A.I.-driven” Alpha School, revealing their internal learning and enrollment data for the first time, just as the network expands into public education. There is a big gap between the marketing and reality. https://www.nytimes.com/...
Alpha's outcomes might be short of what they claim but also better than the vast majority of other schools. There are real questions of selection effects and what it means to learn twice as fast (not what it may seem), as @KelseyTuoc has detailed. https://www.nytimes.com/...
This sounds like the single greatest intervention in the history of education if it were causally true, though - 50% and 80% of students *doubling* learning per year in math/reading?! Proactive, individualized AI for learning makes too much sense. 1/2
The NYT article about Alpha school tried so hard to frame the results as negative even though they seem super impressive. 42% of students progressed 2x as fast as national averages. Among students who started out below median, “only
I'm not Alpha School fan, but framing here is nuts: Only 50% of students grew 2x faster! Even as some defend pub schools that can't grow kids even 1x! And from same author who in 2021 asked “Does It Hurt Children to Measure Pandemic Learning Loss?
I left this comment on the piece, which doesn't seem to be getting approved (though it may just be slow). — I'm the principal at one of Alpha's specialty schools. Regarding 2x learning: 2x is our average and our target. It is true that not all of our students hit that mark. For m…
My two oldest started Alpha in September. It's too early to know if they're learning at 2x speed, but they definitely love school. They complain if we take them out for a dentist appointment or a wedding.
I have a ton of empathy for @Alphaschool skeptics, because I doubted it myself when I joined 10 months ago. I was also one of the louder skeptics about EdTech, all along. My recent piece could have been titled
Alpha School is following the Gartman Hype Cycle. Expectations peaked early 2026. Today, there is increasing skepticism. The important question is not where they are today but where they will be in a few years when they reach the Plateau of Productivity. I've been critical of @Al…
I was quoted in a @nytimes article on @Alphaschool. I'd like to clarify, many parents at Alpha School were concerned about negative self-talk because their kids couldn't finish their work in the 2-hour window, despite heavy pressure to do so. Leaderboards were just one factor.
“only” 42 percent of students progressed TWICE as fast as projected by national averages in both math and reading. wow alpha school sounds like such a failure