I guess shaming educators works to bring them to an abrupt about-face. But I'm not a big fan of shamers. Shame on them.
I’m not saying Sold a Story got the reading science wrong. I’m saying there is something unsatisfying about explaining a decades-long instructional swerve principally by looking backward at who sold what idea to whom.
Because schools didn’t just swerve in reading. They’ve been swerving all over the place.
Differentiation, personalized learning, intervention, small groups, MTSS, adaptive technologies, digital curriculum, data-driven instruction, competency models, and now AI. Maybe educators haven’t spent all these years wandering from one shiny object to another. Maybe they’ve been trying to solve something.
So, why did schools swerve?
Sold a Story gave education a compelling explanation for one of its more consequential wrong turns, investigating how ideas about how children learn to read became embedded in classrooms, teacher preparation and curriculum even though important elements of those ideas conflicted with cognitive-science research. The series followed the theories, instructional practices, influential educators and publishers that helped those ideas spread, and became part of a much larger national reconsideration of reading instruction.[1]
The investigation matters, but reading was hardly the only place education changed direction, and I think the bigger question is what educators were encountering that made so many new models, practices and technologies attractive in the first place.
Maybe the problem wasn’t simply that schools kept getting sold new answers. Maybe the problem they were trying to answer kept getting harder.
Teaching Got More Complicated
Human beings have never learned identically, and establishing that one instructional practice works better than another doesn’t erase differences in prior knowledge, language, ability, circumstances, cognition, readiness or pace. Better practice still encounters different humans.
School leaders think it's all structured this way because of laws, but that is not so. Grade-based-by-age is a myth.
I recently took on one structural dimension of this in What the Supreme Court Actually Required Was Individual Progress (reposted on our site), looking at the tension between an educational system still heavily organized around age-grade placement, common pacing and time-based progression, while the Supreme Court’s Endrew F. standard for students covered by IDEA requires an individualized educational program reasonably calculated to enable progress appropriate in light of the child’s circumstances.[2]
But the entrenched age-grade structure is only one part of a much larger teaching-complexity problem.
Consider what we now expect a teacher to reconcile during something still recognizable as a classroom lesson: curriculum and standards, prior learning, assessment evidence, interventions, accommodations, language differences, pacing, grouping, available instructional time, supplemental resources and the actual learners sitting in front of them.
Viewed that way, many of education’s supposed swerves look less mysterious.
Differentiation, personalized learning, intervention, small groups, adaptive technologies and instructional data systems were all, in different ways, attempts to close the distance between delivering instruction and producing learning.
Yet we kept adding curriculum, assessment, interventions, software, data and instructional choices without proportionally increasing the capacity of the system to reconcile all of it. Somebody still had to put the pieces together, and that somebody was the teacher.
We effectively made the teacher the integration layer of the instructional system, asking one human being to continuously notice, interpret, decide, locate, differentiate, regroup, reassess and adjust while also doing the fundamentally human work of teaching.
No wonder teachers are overwhelmed.
And Now, AI. Darn It.
AI is here, and I think it is going to strip education to its bones along nearly every vector, not because everything schools have been doing is wrong, but because AI lets us ask a question we could not seriously ask before: Which parts of this enormously complicated enterprise actually require a human being to do the work?
Once machines can interpret information, recognize patterns, consider multiple variables, generate alternatives and operate repeatedly across large amounts of bounded information, it becomes difficult not to notice how much teacher capacity is being consumed by orchestration rather than teaching.
Why does the teacher have to manually reconcile what the curriculum expects with what evidence of learning is showing, find the appropriate instructional resource, keep track of who needs additional support, determine which students can usefully work together, remember what has already been tried and continually decide what ought to happen next?
Those questions aren’t really about whether AI can teach. They are about whether teaching has accumulated a great deal of work that no longer necessarily requires the teacher to personally perform it.
That, to me, is the real disruption.
The Swerve Isn’t Over
Education isn’t done swerving, but the next move cannot simply be from one theory of instruction to another, nor should it be from one generation of education technology into a crop of general-purpose AI tools that teachers are expected to learn, prompt and manage on top of everything else.
The more interesting direction is bounded intelligence, where AI is attached to the actual instructional work, knows the environment in which it is operating, and is constrained by the curriculum, information and educational purposes established for it.
That emerging category is what we call Learning Intelligence Technology, or LIT.
It's above the consumer-level of AI that has so many in the education field worried about AI literacy, whether it should be used at all, a huge story line of its own. This is AI but not a chatbot for students. A visible and also invisible behind-the-curriculum agent for goodness. On the ladder rising to school nirvana, it's the next rung up.
Kiddom is expressing LIT as a new category of AI pointed directly at curriculum and teaching complexity, rather than taking a general-purpose AI and simply putting it in front of a teacher.
Curriculum matters because it creates the boundary. LIT isn’t unrestricted AI deciding what a child ought to learn from the universe of knowledge available to a model. It places intelligence inside the High Quality Instructional Material (HQIM) instructional system, where it can work in the context of the curriculum, evidence of learning and the instructional choices available within that environment.
That changes the question from What can a teacher do with AI? to a considerably more consequential one:
What parts of the complexity currently carried by the teacher can now be carried by the instructional system?
So. Much. More. Interesting.
Frankly, I've been a bit bored with the discussions of should-we or shouldn't-we for consumer technologies that schools really can't control and everyone is already using.
More Than Generative AI
Much of today’s education AI conversation is understandably fascinated with generation: make a lesson plan, create a quiz, explain a concept, produce a worksheet, draft a communication. There is real utility in all of that, but generating more instructional material doesn’t necessarily solve teaching complexity and could, ironically, create more of it.
Learning Intelligence Technology is aimed at something deeper, using bounded intelligence within the instructional environment to reconcile curriculum,
evidence, learner needs and available instructional responses, so that achieving more for students does not automatically mean asking more of the teacher.
Nor is this personalized learning all over again, with thirty learners following thirty independent pathways while a teacher somehow supervises thirty instructional universes. The intelligence is in recognizing what matters, detecting patterns, knowing when learners can move together and when a different response is called for, and helping the instructional system act on those differences without converting every action into another manual task for the teacher.
That makes the achievement question and the teacher-capacity question increasingly the same question.
Maybe Schools Were Swerving Around Something
Perhaps the larger lesson from Sold a Story isn’t simply that schools can be sold the wrong educational idea. They can, and that deserves investigation when it happens.[1]
But pull the camera farther back and another story appears.
Schools have spent decades trying to make an increasingly complex instructional enterprise work, layering new practices, technologies, interventions and information onto a structure in which the teacher remained responsible for putting the pieces together.
AI doesn’t make sound instructional practice irrelevant, and it doesn’t invent the differences among learners that schools have always encountered. What it changes is the assumption that all of the complexity surrounding those things must remain human work simply because, historically, there was no practical alternative.
Now there may be.
That is why the next swerve matters. It shouldn’t be another search for the one true theory of how everyone learns, nor another layer of technology piled onto the teacher’s desk. The more consequential possibility is that intelligence moves into the instructional architecture itself, bounded by curriculum and educational intent, carrying complexity that teachers have historically had little choice but to carry themselves.
We’ve spent decades making teachers more sophisticated at orchestrating the system around learning.
Maybe the next era begins when the system becomes sophisticated enough to help orchestrate itself.
References
[1] APM Reports, Sold a Story. Investigative podcast and reporting project on reading instruction and the research concerning how children learn to read.
APM Reports: Sold a Story [features.a...eports.org]
[2] LeiLani Cauthen, What the Supreme Court Actually Required Was Individual Progress… But Schools Build by Birth Year & Whole Group, Learning Counsel, June 10, 2026.
Learning Counsel: What the Supreme Court Actually Required Was Individual Progress… [thelearnin...ounsel.com]
