The Self-Driving School
While California's classrooms drift without compasses — 84% of schools providing fewer than 5 hours of AI training — one visionary educator has drawn up blueprints for a 10X moonshot that could rewrite the operating system of learning itself.
California's schools are on fire — and nobody has a hose.
The numbers are stark. A landmark May 2026 Stanford University report — part of California's Getting Down to Facts III project — has laid bare a crisis hiding in plain sight: the state that birthed ChatGPT, Gemini, and Claude is educating its children in near-total darkness about the technology reshaping civilization.
84% of California schools report AI being actively used in their buildings. Yet only 25% have formal AI policies. A staggering 57% of teachers receive zero AI training and actively want it. The median school offers fewer than five hours of AI professional development per year — less time than a single Netflix binge.
Meanwhile, students have found their own compass. They are using AI with or without permission, without guidance, without ethical grounding. Lower-income students and students of color — often portrayed as digitally disadvantaged — are actually the heaviest users. The technology has democratized itself. The schools have not caught up.
"The picture is one of minimal proactive engagement with AI for schools," writes Stanford researcher Victor R. Lee, "one that seems to be out of step with the state of California's image as the nation's technology sector leader."
"We do not teach 'AI literacy' as a separate, check-the-box class. Students are treated as junior principal investigators from day one. The workspace is theirs. The moonshot is real." — Dr. Freedom Cheteni · Founder, School of AI (sof.ai)
Marco Troper · 2005–2024
The spark that became a fire
He didn't wait for permission.
He just started building.
Before there was a manifesto, before there was a school, before there was a philosophy of self-driven learning — there was Marco Troper. Young, relentless, and armed with a conviction that would have sounded absurd anywhere outside a garage or a laboratory: biology is fundamentally applied mathematics.
Marco was working on something he called Nanotro — a vision of programmable lipid nanoparticles capable of crossing the blood-brain barrier to target cancer cells with surgical precision. He had no PhD. He had no institution. He had first principles, extraordinary courage, and the belief that if you could map the mathematics of living systems, you could write the software to heal them.
It was Marco's spirit — his refusal to wait for gatekeepers, his absolute conviction that learning must be self-driving — that became the philosophical core of Dr. Cheteni's School of AI. "Marco understood something that takes most people decades to accept," Dr. Cheteni has said. "The moment you accept that knowledge has borders, you stop building. Marco never accepted that."
At sof.ai, every student who arrives to work on a moonshot project carries a piece of Marco's torch. The Nanotro challenge — given to new students as their first assignment — asks them to solve the three engineering questions Marco was wrestling with: the address, the engine, and the off-switch of a cancer-targeting nanoparticle. Not as homework. As real science.
Marco would have been 27 this year. His project continues.
The gap between what is and what must be
Of California schools report AI being actively used — yet guidance, policy and training have not kept pace.
Vast majority of CA schools have no ethical AI use policy. Students and teachers are navigating alone.
Only 6 of 81 California principals report formal AI literacy instruction in their schools. Elementary: 4%.
Dr. Cheteni proposes 100-hour structured educator research residencies — against a national median of under 5 hours.
Average number of AI tools used by teachers in AI-supportive CA district — including tools their district doesn't provide.
Black and Hispanic teens use AI for schoolwork at higher rates than White peers (52%). The future is already here — unequally.
Data sources: Stanford University / Getting Down to Facts III (Lee, 2026) · RAND Corporation (Doss et al., 2025) · Pew Research Center (McClain et al., 2026) · CoSN (Maylahn, 2025) · Code.org et al. (2025)
Four Pillars of the
Self-Driving School
Dr. Cheteni's response to the crisis is not incremental. It is a complete reimagining — built on four load-bearing ideas that upend every assumption of traditional education.
De-Center the Classroom
Replace the lecture hall with the Autonomous Research Studio. Inspired by School 42's teacher-free model and the TRICK framework (Trust, Respect, Independence, Collaboration, Kindness), students are treated as junior principal investigators from day one. Disciplinary boundaries — between math, biology, ethics, and code — are dissolved entirely.
Eradicate "Shadow AI"
Every student receives a secure, local, open-source AI agent ecosystem. AI tools are trained not to give answers but to act as Socratic thought partners — simulating environments, asking the questions that cause students to discover. Rather than banning tools or using faulty plagiarism detectors, sof.ai makes ethical use the baseline of how learning happens.
100 Hours, Not 5
Educator development is not an annual presentation — it is a paid active research residency. Teachers are given 100 hours of structured time to co-design AI-driven learning tools alongside engineers. A network of Teachers on Special Assignment (TOSAs) leads ongoing scientific inquiry within their schools, not PowerPoint presentations.
Ubuntu in Action
Built on the African philosophy of Ubuntu — I am because we are — every breakthrough at sof.ai is open-source and free for any public school. Code, curriculum, and nanotech simulations: all shared. Students learn the Spinoff Model, ensuring any healthcare breakthrough remains accessible to the communities that need it most.
Why California can't
afford to wait
The Stanford GDTF III report — perhaps the most comprehensive assessment of AI readiness in K-12 education ever conducted for a single state — lands like a verdict. California, home to OpenAI, Google DeepMind, and Anthropic, has produced a generation of AI companies without producing a generation of AI-literate citizens.
The irony is not lost on Dr. Cheteni. "We have built the engines of transformation," he says, "and handed the keys to children with no driver's education." The state that shaped the internet is, by its own data, doing less than almost any other state to prepare its students for the world the internet is becoming.
The data tells a story of remarkable, quiet democratization. Minority students and lower-income households are already using AI chatbots for schoolwork at rates that exceed their more affluent peers. This is not a problem; it is a signal. These students are self-teaching, adapting, finding their way to tools that offer leverage. What they lack is not motivation — it is guided excellence.
Dr. Lee's recommendations in the GDTF III report call for minimum professional development hours, state funding, enhanced policy support, and committed AI literacy integration across K-12 — from elementary school onward. They are urgent, evidence-based, and eminently reasonable.
Dr. Cheteni's response is: all of that, times ten, and then some. Not because moderation is wrong, but because the moment calls for something the education system has never truly seen: a school designed from the ground up for the intelligence age.
The moonshot
is the curriculum.
At sof.ai, the opening assignment for every student is not a syllabus review or an icebreaker exercise. It is an engineering problem that has never been solved. Inspired by Marco Troper's unfinished work, the Nanotro Challenge asks students to answer three questions through first-principles research, AI simulation, and collaboration.
The AI does not give answers. It asks better questions. The workspace is yours.
How does Nanotro find a cancer cell without harming a healthy one?
What is the mathematical "biomarker" address? Students must derive the specificity problem from cellular biology, statistics, and computational geometry.
How does the nanoparticle move through a living body?
Chemical gradients? Magnetic fields? Biological receptor locks? Students choose their propulsion mechanism and must mathematically defend the tradeoffs.
Once it arrives at the tumor — how does it deliver the cure?
Payload release mechanisms, membrane permeability, therapeutic dosing — students design the terminal delivery system and stress-test it with simulation.
The workspace is yours.
California's learning crisis is real, documented, and urgent. But so is the vision of what school could become when learning is self-driving, educators are research partners, and every student is a principal investigator. The torch is lit. The only question is who picks it up.
Begin Your Moonshot
This article synthesizes findings from the Getting Down to Facts III project
(Victor R. Lee, Stanford University, May 2026) with the educational vision of Dr. Freedom Cheteni.