In today’s Diary:
K-12 by age 10
AI = the great equalizer
The 9-year-old who runs a food truck
Intellectual weight-training for kids
Your kids need you to figure this out
Dear Rational Optimist,
I’m officially a resident of Abu Dhabi, having moved my family here last week from Ireland.
Of the many reasons we moved, #1 is:
I am determined to give my kids the best possible education.
So when I tell you I am obsessed with the topic of today’s Diary – how AI has triggered the biggest disruption to education since the printing press – I’m not just saying it. I’m living it!
I loathed school. For me, one-size-fits-all classrooms ground the joy out of learning. I could understand making kids suffer through this if it were producing great results.
It is not.
In the US, math scores have been falling since 2013. Reading scores have hit their lowest levels since 1992. More 8th graders than ever before can barely read.
This is an emergency.
We’ve long known how to give kids a world-class education. In 1984, University of Chicago professor Benjamin Bloom confirmed what aristocrats knew for centuries: the best way to teach is with a dedicated, one-on-one tutor who can personally tailor each lesson.
Bloom found just seven weeks of personalized tutoring could transform C students into straight-A students. From average to exceptional.
Critically, almost any student had the potential to achieve top grades. But very few parents can afford personal tutors. So conveyor-belt schools were the best we could do. Until now.
What happens when the cost of a personal tutor falls to almost zero? We’re about to find out.
A quick story from ROS co-founder Dan Steinhart about his 7-year-old son:
1st grade seemed to be going well.
Jack was in the highest reading group all year. Then we got his test scores. Math was good, reading was bad. He was reading at the level of an early Kindergarten student.
My first instinct was to call the teacher. Instead I did what a good Rational Optimist should and looked for solutions. I tried out four different AI reading programs and settled on one called Ello.
Jack reads aloud with it for 20 minutes every morning on an iPad. It listens to him, gently corrects his mistakes, and only moves ahead once he’s mastered a concept. I do it with him when I can, but I don’t have to. It’s gamified, rewarding him with tokens for completing a lesson, so he loves it.
In seven weeks, he has jumped 1.5 grades in reading level.
For $15/month.
And this is the worst it’ll ever be.
Many folks are understandably skeptical about their kids staring at an iPad to learn.
“Stephen, where’s the hard proof AI schooling works?”
Oh, we’ve got proof.
The World Bank recently studied 422 Nigerian high school students. These students interacted with ChatGPT to boost their English grammar, vocabulary and writing skills over a dozen 90-minute sessions.
The students made learning gains equivalent to two years of traditional schooling – in six weeks! Girls and struggling students saw the largest gains.
AI = the great equalizer.
At the other end of the socioeconomic spectrum, look at Alpha School, the $40,000 a year private AI-first school in Austin, TX.
When we visited the University of Austin (UATX), two different parents raved about it.
One was a well-known investor with two kids at Alpha. His 1st grader jumped from the middle of the pack to the 99th percentile in reading and 94th percentile in math. “I don't know if I could put my kids back in a normal school ever again and feel good about it,” he says.
Alpha has totally reimagined the school day. There’s a dozen or so students per class, but each kid follows a personalized, AI-crafted learning plan. They go at their own pace and can’t progress to the next step until they pass a mastery test.
This is possible because an AI tutor assesses their knowledge in real time, identifying gaps and helping them through it.
Alpha’s students learn, on average, 2.6X faster than other kids. That’s the equivalent of finishing K–12 academics before age 10. They also score in the top 1% in every class and every subject.
Here’s the real kicker: they’re accomplishing this in a two-hour daily sprint dedicated to core academics. With no homework.
They spend the rest of the day practicing human skills. Public speaking, riding bikes, launching podcasts, building businesses. One dad told me his 9-year-old son learned how to run a food truck.
This gets at a big misconception about AI learning:
This is not kids staring at iPads all day.
AI learning works so well, it compresses all the core academics into two intense hours.
Then the kids get the entire rest of their day to pursue their curiosities and DO stuff.
Rational Optimists who read our AI guide know the No. 1 skill for the AI era is: Be more human.
Creativity, problem solving, empathy and the capacity to inspire others will become the new “hard skills.” Alpha isn't just teaching kids; it's cultivating humans.
Imagine when these students hit high school age.
MacKenzie Price launched Alpha School after her second-grade daughter said, “School is so boring.”
Here’s MacKenzie: “In two and a half years, they had taken a child who was tailor-made to love school and be curious and interested, and they wiped away that passion.”
This mirrors my own motivations. School crushed my soul, and I want better for my kids.
What about teachers?
Alpha’s model makes teachers more important. In regular schools, teaching is an impossible job. One teacher scrambling to teach 25 kids with wildly different abilities. Some kids get left behind while others get held back.
Meanwhile, teachers are bogged down grading papers, designing lesson plans and managing behavioral problems.
At Alpha, teachers are called “guides.” They don’t spend their time directly lecturing or grading assignments. The AI does that. This frees up guides to do what they do best: inspire, mentor and motivate students.
Elevating teachers from human box checkers to mentors is a brilliant example of one of my core beliefs: AI will handle the grunt work so we can focus on more important stuff.
I recently chatted with Alpha School co-founder Jonathon Stewart on the ROS podcast. Click here to watch it.
And if you're interested in bringing an Alpha School to your own town or city, register your interest here. They already have six locations across America and are looking to expand to locations where 10 or more families register. Do it!
Another AI-first school every Rational Optimist should know about:
Primer is building a network of "microschools" – small, in-person learning centers run by passionate teachers who think more like entrepreneurs.
At Primer, the morning is dedicated to core academics. Students use laptops to move through personalized math and reading tracks at their own pace. The AI adapts to each child’s level in real time.
The afternoon is for "Pursuits," where students dive into passion projects like starting a Shopify store for slime or creating a YouTube science series. Students are encouraged to use AI to generate ideas, branding, and business plans.
Primer is what happens when you build school around kids’ individual curiosity. It’s possible to start your own microschool. Check it out.
You can tap AI without a special school.
Synthesis is an interactive math tutor, born out of the experimental school Elon Musk created for his kids at SpaceX. It’s open to everyone, everywhere.
Instead of drilling kids with facts or multiple-choice questions, Synthesis throws them into fast-paced simulations with no easy answers. Students might redesign a failing Mars colony, negotiate a trade deal, or debate how to allocate scarce resources. They have to strategize and explain their logic on the fly.
Synthesis is designed to teach students how to think. It runs on AI models trained to mimic world-class teachers and push students to concentrate on hard problems.
As Synthesis co-founder Chrisman Frank told me “The point isn’t just solving the problem, it’s how you solve it. How long can you stay with the hard thing?
Synthesis pushes the boundaries of what school even is. It’s the closest I’ve seen to intellectual weight training for kids. I’m testing out Synthesis with my six-year-old this summer. Check it out here.
Khanmigo is Khan Academy’s AI-powered tutor, built in partnership with OpenAI. For students, Khanmigo is an infinitely patient guide who asks probing questions to help them arrive at the answer independently.
Khanmigo is also a revolutionary tool for teachers (it’s free for US teachers, courtesy of Microsoft). Without it, they spend 40–60% of their time on admin and prep. It acts as a tireless teaching assistant, capable of drafting lesson plans in minutes and providing personalized feedback on student work.
This frees up educators to answer questions and mentor their students. Maybe that’ll help reignite the spark of curiosity in the children it touches. Access Khanmigo here.
Above all else… don’t wait. Mass market elite tutoring is here.
Your kids or grandkids need you to figure this out and show them.
Those who do will have a massive head start. Some will call it an unfair advantage.
But Synthesis costs $99 per year for up to 7 kids. Ello is $15/month. Khanmigo is $4/month.
So really, it comes down to parents who’ll take charge.
And we haven’t even touched on one of the most exciting AI strategies yet: using ChatGPT to create your own personal AI tutor. More on that next week, including how I’m using AI to better educate my own kids.
Until then…
Please click those “Like” and “Restack” buttons below.
And when you test out these AI platforms, leave us a note in the Comments section below letting us know how it went.
—Stephen McBride
Stephen McBride is a co-founder of the Rational Optimist Society.
Kids needs more than an ",expert" AI tutor.They also need motivation, focused encouragement,possibly mentoring for some tasks or even correction or disciplining and not just some sycophantic AI tutor.
I like the idea of using AI to accelerate learning, as there's more to absorb now than at any time. So, I'm with you on new tech for schooling. I am skeptical of the current tool suite, however. My scatterings off AI (ChatGPT and various other LLMs) have been less than useful in improving my work and my learning. I don't think we're there yet - particularly in math and empirical science. There was a recent editorial that indicated at the college level, the AI tools were actually helping students avoid the work they precisely needed to do in order to think and write properly. This is a pitfall that needs to be managed at other levels as well. And the training sets for LLM's include all manner of "below average" inputs and it generates mediocre results because of that. So, what we select as training materials for educational AI's need some careful thought and curation.