A San Francisco venture capitalist just decided a chatbot could do a better job teaching his five-year-old than a classroom of trained teachers could. He’s paying $75,000 a year to find out if he’s right.
That’s the real story behind a wave of coverage in July 2026 about Silicon Valley’s newest status symbol: AI-run private schools promising to reinvent the school day. It’s easy to read stories like this and file them under “rich people problems,” then close the tab. But this one is worth sitting with. It’s really a story about AI homeschool tools moving down-market fast, and about what happens when the people with the most money get first access to something nobody has fully proven yet.
Every other take on this story asks whether it’s ethical to let AI teach a kid. That’s a fair question, but it goes in circles fast. It also skips the more useful question underneath, the one most parents actually want answered. The sharper question is about access. The wealthy have already placed a six-figure bet that AI tutoring works. So what does that mean for the version of that same idea sitting in an app on your phone right now, for a fraction of the price?
Here’s the honest twist before you get too worried or too excited. The best evidence that AI tutoring beats a classroom comes from a study of college physics students, not five-year-olds. Even the families writing six-figure tuition checks are buying a bet, not a guarantee. Whether or not you’d ever consider a five-figure kindergarten tuition, that gap between the hype and the evidence is worth understanding before you decide what this trend actually means for your own family.
Wealthy Parents Are Already Paying To Let AI Teach Their Kids
Some of America’s wealthiest families have stopped asking whether AI belongs in the classroom. They’ve started writing checks instead. The Verge reported in July 2026 that companies like Forge Prep and Alpha School charge families tens of thousands of dollars a year for AI tutoring and “interactive project-based workshops.” The outlet had a blunt name for what the kids actually get: “beta testers for unproven tech.”
Shaun Johnson, a San Francisco-based venture capitalist, told The Wall Street Journal he plans to enroll his son in a $75,000-a-year “Alpha Kindergarten.” His reasoning is worth reading in full: “We recognize that education is likely broken the way it is and there’s going to be entrepreneurs that try to fix it… You want someone to be able to think on their feet and navigate the world, not necessarily a recitation of facts in a particular discipline.”
Alpha School co-founder MacKenzie Price started the school in 2014 alongside co-founder Brian Holtz. She’s said she plans to keep “hot-button social issues” out of the AI-driven curriculum, a detail that’s easy to skip past but worth holding onto. Hedge fund manager Bill Ackman has become one of the school’s loudest public boosters, calling it on social media a breakthrough he hasn’t seen in K-12 education in years. That kind of public praise from a well-known investor tends to travel fast among other wealthy parents weighing the same decision.
None of these companies publish independent performance data. That means every family enrolling right now is taking Alpha School’s word for it. The wealthy aren’t asking if AI tutoring works anymore. They’ve already answered that question with their checkbook. The more interesting question is what that answer means for everyone who isn’t spending six figures a year to get it.
What $75,000 A Year In AI Education Actually Buys
Strip away the branding and Alpha School’s core idea is fairly simple. Students spend the first couple of hours of the school day on a highly personalized, AI powered curriculum covering core academics like math, reading, and science. This kind of personalized learning tracks each child’s learning in real time, so a student moves at their own pace instead of the class’s pace. The system, built on a platform called 2 Hour Learning, is designed to compress traditional instruction time. Kids move faster through material they’ve already mastered and slow down where they actually need help. The pitch, in the company’s own words, is giving kids their time back, freeing up the rest of the school day for the life skills work traditional schools rarely have room for.
That remaining time shifts to project-based work: public speaking, entrepreneurship, and financial literacy, taught by adults acting more like mentors and coaches than traditional teachers delivering lessons from the front of a room. The school’s principal and primary funder, Joe Liemandt, built his fortune in automation and enterprise software as founder of Trilogy Software, becoming one of the youngest members of the Forbes 400 in the process. That background shaped the company’s pitch that education can be streamlined the same way software gets streamlined. It’s a compelling pitch, and one reason wealthy early adopters have been so quick to buy in without waiting for outside proof.
That pitch has a real crack in it. A 404 Media investigation found that Alpha’s AI-generated lesson plans were internally flagged as sometimes doing “more harm than good.” The same investigation found the underlying platform had scraped training data from other education platforms, including Khan Academy, without permission. Alpha School disputes the characterization as inaccurate. Either way, the product being sold here isn’t a new invention. It’s a repackaging of AI-powered personalization that already exists elsewhere in cheaper, more accessible forms, wrapped in a premium in-person experience with a Silicon Valley pedigree.
What The Research Actually Says About AI Tutoring
The strongest evidence that AI tutoring works comes from Harvard, not from a kindergarten. A 2025 randomized controlled trial split students into groups at random, so the results can’t be chalked up to who signed up. Physics lecturer Gregory Kestin led the study, published in Scientific Reports, testing an AI tutor called PS2 Pal against in-class active learning with 194 undergraduate physics students. Students working with the AI tutor posted a median score of 4.5 against a baseline of 2.75. The classroom group scored 3.5. That’s a large, statistically significant effect, with researchers reporting a range of 0.73 to 1.3 standard deviations, that Harvard researchers described to The Hechinger Report as learning more than twice as much in less time.
Brookings’ own review of the research points the same direction. A UK trial found a pedagogically fine-tuned AI tutor helped secondary math students transfer knowledge to new problems better than static hints did. A World Bank study in Nigeria found an AI chatbot tutor helped first-year secondary students gain the equivalent of one and a half to two years of typical schooling, in just six weeks, at a cost of roughly $48 per student for the pilot. A Stanford study of 900 tutors and 1,800 K-12 students found an AI copilot lifted math mastery across the board, with the biggest gains going to lower-rated, less experienced tutors. Brookings’ synthesis of this research is genuinely encouraging about what well-designed artificial intelligence tutoring can accomplish in a real learning experience.
That’s the honest caveat worth repeating from the top: every one of these studies tested teenagers, college students, or working adults. None tested five-year-olds. Even the rich aren’t buying a fully proven product for the age group they’re deploying it on. They’re buying a bet, same as anyone else would be.
The Access Gap Nobody Is Talking About
Here’s the direct claim underneath this whole story: the AI capability isn’t the scarce resource. The packaging is. A $75,000-a-year price tag buys curation, a physical campus, and a curated peer group. It does not buy a fundamentally different AI tutor than what’s already sitting in far cheaper, unbundled tools available to any family with a laptop and a decent internet connection.
MacKenzie Price’s decision to keep “hot-button social issues” out of the curriculum is a real trade-off worth naming plainly, not an accusation. It’s a curated, privately designed choice that most public schools, and even most private ones, simply don’t have the room to make. Public school curriculum goes through school boards and elected oversight; a homeschool curriculum built around one company’s AI answers to neither. Meanwhile, a year-long Brookings Global Task Force study gathered input from more than 500 students, teachers, and parents worldwide. It concluded that, at this stage, the risks of AI in education still outweigh its benefits when tools aren’t deployed thoughtfully, especially around data privacy and over reliance on AI at the expense of critical thinking. That’s a genuinely sobering finding from researchers who are otherwise optimistic about the technology.
None of the AI-driven schools named in this story publish outcome data. That means the earliest adopters, disproportionately wealthy and well-connected families, are effectively the pilot group for a product the rest of the market will eventually judge by their results. When the best-resourced families get first access to something unproven but promising, the risk and the reward of being early both land on them first. Everyone else inherits whatever they learn, for better or worse, often years later and without the same safety net.
That’s not a new pattern. It’s how most new technology reaches ordinary families, from the home computer to the smartphone, and now the AI tutor. The people with the most money and the highest risk tolerance go first, and what they learn eventually filters down to everyone else.
How To Think About AI Tutoring Without The Elite Price Tag
The real question was never whether letting AI teach your kid is a good idea in the abstract. It’s what a responsible, age-appropriate version of what the wealthy are already testing would look like, without the $75,000 price tag or the risk of using it on an age group nobody’s actually studied yet.
The research is fairly consistent on this point: AI tutoring works best as a supplement, not a replacement for human connection. Free and low-cost resources for building your child’s own learning experience already exist, and most cost nothing close to a private school tuition. Many homeschool parents already use AI tools like Khan Academy’s AI-powered tutor to let a child work through core academics at their own pace, and it’s a genuinely valuable tool when it’s paired with, not swapped in for, real adult attention. In practice, that might look like using an AI tutoring app for extra math practice after homework is already done, not instead of it, or letting a chatbot re-explain a concept when the first explanation from a teacher or parent didn’t land. A parent still stays involved in the parts AI can’t replace: social-emotional growth, digital literacy, and honest conversations about hard topics a curated curriculum might quietly skip.
Homeschool families integrating AI into their curriculum should treat screen time the way they’d treat any other homeschooling decision. Keep it purposeful and time-boxed, use appropriate resources for your child’s age, and balance AI powered learning against hands on projects, creative projects, and offline activities that build real world skills and problem solving no chatbot can teach. Responsible use, not blanket avoidance or blind trust, is what separates a helpful AI tutor from an over reliance that crowds out a child’s own thinking.
The Bet Any Family Can Make
You don’t need six figures of tuition to test whether this technology has something real to offer your child. You need the same honest curiosity the wealthy are paying to indulge, minus the price tag and the pressure to prove a five-year-old is worth the investment. Plenty of parents already reach for AI tutoring tools without spending a dime on a boutique school.
That’s genuinely good news. The underlying capability driving Alpha Kindergarten isn’t locked behind a velvet rope. It shows up in far cheaper, more accessible AI tutoring app options already built for exactly this kind of supplement, not substitute, use. If you want a place to start looking, Nollytech’s guide to the best K through 12 learning platforms is built around tools ordinary families can actually access.
Try one. Watch how your kid responds. Ask yourself the same question Shaun Johnson is paying $75,000 a year to answer: is this making my child a better thinker, or does it just look like it? Nobody has that answer yet, rich or not. That’s exactly why it’s worth finding out for yourself, on your own terms, with your own judgment doing the work an AI tutor was never built to replace.
