What Has Changed and What Hasn’t Changed in Education

What Has Changed

AI is dramatically better than it was a year ago. New models have emerged that are not just incrementally improved but fundamentally more capable.

That part is clear. What hasn’t changed requires a longer conversation.

What Has Not Changed

• Direct Instruction Remains Critical for Improving Student Outcomes

Direct Instruction in the Engelmann tradition remains the most effective method of teaching we have — not because the evidence is beyond dispute, but because it follows the discipline of engineering: analyze, design, field-test, evaluate, revise — the same cycle that produces reliable software, bridges, and aircraft. Engelmann did not start with ideology and then try to make kids fit it. He started with a logical analysis of the stimuli and the kids’ behavior — their responses, their performance, their errors — and kept iterating until they learned. Cognitive science and the broader science of learning have spent more than 50 years confirming what this process produced, through both direct and indirect research. Whether by analog or by digital, by script or by screen, we must teach students effectively and efficiently, using principles of learning science and instructional design.

• Most Schools Lack a Coherent Curriculum

However, most schools don’t even have an articulated, sequenced, cumulative curriculum. What they have is at most a patchwork, or something that lives in the head of individual teachers. It leads to teaching that is built on sand — it is almost irrelevant whether a teacher teaches explicitly or not when what is taught is random slop that was Googled or generated by AI the night before. And it all just keeps moving. A student who failed to master a prerequisite skill three months ago has no mechanism to recover, because the curriculum doesn’t build or circle back. There is also zero transparency — as a parent or a consultant, trying to find out what is being taught in a given grade for a given subject is like going on a scavenger hunt, where at every turn you’re told to just trust the process. This isn’t for kids. This is because the adults can’t get their act together.

• Meeting Various Student Needs Involves Compromises

Even with a coherent curriculum, meeting every student’s needs is structurally difficult — if not impossible — in the heterogeneous classroom. Learning is like a stepladder: mastering prerequisite skills enables more complex skills. Cognitive science compels us to respect the expertise reversal effect: what helps a novice overwhelms an expert, and what challenges an expert leaves a novice behind. We must also be empathetic to the fact that some students require many trials to learn something, and others very few. In a class of thirty, that means thirty different calibrations. No teacher on earth can provide them. You cannot teach one child to decode words while teaching another to interrogate Lady Macbeth’s motives. Someone always loses — the highs who are bored out of their skulls, the lows who are found coloring in the back of the room, and all the kids who just need a little more time, or who came to school already knowing the content. No amount of skill, effort, or calls to “differentiate more” changes the math.

• The Quality of Teaching Is Dramatically Different from Classroom to Classroom

Every school has its exceptional teachers — the ones parents request by name, the ones whose students remember them decades later. But they are in thin supply. The rest of the staff in a building span an enormous range. Some are brand new and learning through trial and error. Some are mediocre and have plateaued. Some have been teaching for 30 years but never developed beyond the level of someone walking into a classroom for the first time. And some have quietly given up — going through the motions, counting the years to retirement, while a room full of children sits in front of them. This is not entirely their fault — the system that trained them, placed them, and left them unsupported bears much of the blame. But the children in those classrooms experience the consequences all the same, and we owe it to them to be honest about that. To pretend this doesn’t happen is to look away from the students trapped in those classrooms. It is to prioritize adult narratives over children.

• Teacher Prep Produces Ideology, Not Effective Teachers

Part of the reason teaching quality varies so wildly is that we never trained teachers to teach in the first place. Education schools are not in the business of producing skilled instructors. They are in the business of producing adherents to progressive education — taught by professors who have weird ideas about teaching and who enjoy the luxury of never having to face a classroom of 30 kids. A new teacher can write a reflection on equity in the classroom but cannot deliver a clear, efficient lesson on two-digit subtraction. The gap between what ed schools teach and what children need is not a minor oversight. It is a systemic failure that reproduces itself every year, in every cohort, across every state.

• The Educational Lottery Persists

For parents, the variance in teaching quality amounts to an educational lottery. On any given year, the draw of a hat determines whether your child gets the teacher who ruins their year or the one down the hall who would have transformed their academics and their self-concept. The most engaged parents trade notes in parking lots, jockey for specific classrooms, and dread the assignment letter each fall. The parents without the time, connections, or confidence to navigate the system get whatever is left. The system treats this as normal. It is not normal. It is negligence, and we owe it to children that they can all have the entitlement of a highly effective teacher, regardless of zip code or parental involvement.

• There Is No Accountability and Nearly Unlimited Autonomy

Once a teacher closes their classroom door in many schools, they are provided a level of autonomy and lack of accountability that is essentially unthinkable in other professions. A surgeon does not get to invent their own technique, refuse to track patient outcomes, and never be audited. But in education, teachers are handed a room full of children and left alone to do whatever they believe is best — regardless of whether it works. At most, teachers receive a handful of formal observations per year — brief, judgemental, and largely performative, using a tool like the Danielson Framework that actually undermines direct, explicit teaching. Outside of those moments, no one is checking whether students are learning, whether the lesson makes sense, or whether the teacher is even teaching at all. Many teachers have been led to believe they have already arrived at the top of the mountain, and the Dunning-Kruger effect is rife. All attempts to change this in education seem to fail.

• Teachers Are Not Going Anywhere

There is a recurring fear that AI will replace teachers. It won’t — for a reason so obvious it often gets overlooked: parents need to go to work, and their children need somewhere to be. There will always be buildings full of children who need to be supervised, supported, and cared for by adults. But here is the harder truth: the current model is already failing on its own terms. The U.S. has roughly 400,000 teaching positions that are either vacant or filled by someone not fully certified for the role. Over half of teachers report chronic burnout, and about one in three say they are likely to leave the profession within two years. We are asking teachers to do something that is, as others have argued, structurally impossible — and then acting surprised when they break under the weight of it.

What may change in the age of AI is a restructuring of the old job description. If AI takes over the design and delivery of instruction, then the adults in the building take on something different. Call them guides, coaches, facilitators — call them whatever you want. Their role shifts from content delivery to motivating students, building relationships, and ensuring every child is actually engaging with the material in front of them. Teachers are not going anywhere. But teaching might finally become a job a human being can actually do well.

What Has Changed for Me

It’s no secret that I’ve started working on the Timeback platform as a Chief Learning Scientist,  joining Carl Hendrick, Sarah Cottingham, and Becky Allen in that role. Every reason I’ve just outlined in this post is why. I have spent my career observing classrooms, consulting with schools, and watching the same structural and systemic problems repeat themselves. Teachers deserve better. Students deserve instruction that actually meets them where they are. Parents deserve transparency and confidence that their child’s education isn’t left to chance.

Timeback is my latest attempt to stop describing these problems and start solving them. The platform is being infused with the Direct Instruction principles I’ve spent my career advocating for: sequenced, mastery-based programming grounded in cognitive science, delivered with the precision and personalization that only technology can provide at scale — while keeping humans at the center of the experience, doing what humans do best. This is not about bowing to a dystopian future, nor handing education over to Silicon Valley, nor innovation for innovation’s sake. It is about the quest to finally grant every child the right to reach their limitless potential.

Skeptics have pointed out that this has started in private schools as though they’ve uncovered a fatal flaw. They begin by referring to the current cost of private schooling, exposing their ignorance of the fact that every transformative technology started expensive. Cell phones, air travel, flat-screen televisions — all were luxuries before they became ubiquitous. The trajectory from exclusive to accessible is not a flaw in the model, nor in the character of those who choose to involve themselves in this project. It is how the world has always worked. A little more historical awareness, and a little more generosity in interpreting the intentions of others, would go a long way.

Instead, I ask readers to consider a few questions. Could a model like Timeback help children access world-class instruction regardless of their circumstances? What about the kid who wants to fly faster? The kid in an alternative school who needs to go back and fill gaps caused by chronic absenteeism? The child in a rural district where there is no AP physics teacher — and never will be? The student with a disability whose IEP goals haven’t changed in three years because no one knows how to move them forward? The parent clamoring for Direct Instruction in a district that is beholden to the malpractice of Discovery Learning, Crayola Curriculum, and Building Sinking Classrooms? The school performing at near single-digit proficiency with nowhere to go but up?

This is the whole point of the work. But it has to be done responsibly. I have been troubled by recent cynics calling for this model to be rushed into the most vulnerable classrooms. We do not test new medicines on the sickest patients before we know they work. We establish safety and efficacy first, then expand access. Education technology should follow the same logic. Once our apps are fully designed to embed every principle of learning science and Direct Instruction — and have been rigorously tested — I look forward to bringing a proven product to the children who need it most.