What We Lose When the System Replaces the Teacher
Teaching is being dismantled as a profession. The pitch is optimization. The evidence is thin. The instruction is broken. And the children at the center are being taught by systems built from practices the field already knows don’t work.
This book is about what we lose when we decide that teaching a child is a problem to be optimized rather than a relationship to be honored.
A private school in Austin is the model. The teacher has been engineered out of the room. In her place, an uncredentialed contractor called a Guide, hired through a staffing company, paid well, given no authority to teach. Apps run the instruction. None have been independently validated. The school claims extraordinary results. None have been independently verified.
The pattern predates education. The people building this model made their fortunes replacing skilled professionals with contractors managed by software. Education is the latest industry on the list. The US Secretary of Education has publicly endorsed the approach. Seven state charter boards have rejected it.
The author tested every major AI model on grade-level reading prompts. The vast majority defaulted to practices that reading science spent decades proving don’t work. The science won the argument. The wrong side won the training data. Still Human follows the money, audits the evidence, and confronts the movement to replace teachers with technology that does not know how to teach.
If your child is in school, the tools arriving in the classroom were not built for learning. They were built for scale. Still Human is the reported investigation into what that trade is costing, and what to look for before it reaches your kid.
You have watched the profession be redefined around you. You know what the apps cannot do. This is the case on the record — evidence, source documents, and the argument that what a teacher does is the thing the replacement cannot replicate.
Forty states passed Science of Reading legislation. The AI tools procured under those same states are bypassing it. Still Human names the gap, the procurement processes that missed it, and the decision points that remain open — in terms that hold up in a hearing.
Sold a Story (Hanford/APM, 2022) exposed the reading methods. The Digital Delusion (Horvath, 2025) proved screens harm learning. The Anxious Generation (Haidt, 2024) documented the mental health crisis. Each assumed the teacher was still in the room.
Still Human asks the question none of them addressed: what happens when someone uses AI to remove the teacher entirely — and the AI at the center of it absorbed the wrong methods? The audience is every parent with a child in school, every teacher watching their profession be redefined, and every policymaker who voted for Science of Reading laws without realizing the AI tools are already bypassing them.
The US Secretary of Education has publicly endorsed a school model rejected by seven state charter boards. Wired and 404 Media are publishing on documented student distress inside AI-first classrooms. Forty states have Science of Reading legislation, and those same states are spending tens of millions on AI reading tools adopted through procurement processes that never evaluated the pedagogy. Still Human connects the dots.
Alpha School is not a section. It is a thread woven through every chapter. New revelations surface throughout, each time expanding the scope from a single school to a national strategy.
Cold open. A room full of six-year-olds learning to read from screens. The adult in the room is not a teacher. She is not allowed to teach.
A real classroom. A first-grade teacher catches something no app on earth can catch. The reader falls in love with what’s at stake before the book shows what’s being destroyed.
Joe Liemandt’s career as a pattern. The same play every time: replace skilled professionals with contractors managed by software. This time, the professionals are the last people in the room who might catch the error.
Every major AI model was asked to write a reading lesson. The vast majority got it wrong in the same way, for the same reason. The science won the argument. The wrong side won the training data.
Teachers who retrained in the Science of Reading can catch the AI’s errors in real time. They are the only firewall. The model that replaces them doesn’t need them. That’s the point.
Alpha claims extraordinary learning gains. The test they use doesn’t measure what reading science says matters most. Their principal fed student data into a chatbot and published the output as evidence.
Alpha isn’t an outlier. States are spending tens of millions on AI reading tools whose evidence doesn’t survive a close read. The procurement gap is structural, not accidental.
Opaque LLCs. A Delaware entity incorporated one day before a major donation. A child connected to leadership sitting in the gallery at the State of the Union. Follow the money.
Extrapolate. The new inequality isn’t access to technology. It’s access to a teacher. The schools that kept human beings in the room become the luxury option. Everything else is an app and a Guide.
The steel-man chapter. One builder got it right. Proof that doing it right is possible. Proof that the shortcuts miss everything.
Return to the real classroom. The teacher is still there. A qualified human being in a classroom is not a luxury. It is a right.
Josh Durey spent twenty years in international education, from first-grade classrooms to school leadership, before leaving to investigate what AI is doing to reading instruction. He retrained himself in the Science of Reading after recognizing that the methods he had been taught, praised for, and promoted through were debunked by the research.
That experience gave him both the classroom expertise to evaluate AI reading instruction and the personal stake to investigate where it’s heading. He designed the evaluation instrument at the heart of the book’s original research. He built the AI-powered curriculum pipeline that funds the project. He lives in Seoul with his wife and two daughters.