Author: Jeffrey Stewart

The Promise We Thought We’d Have by Now

If you’d told me five years ago that artificial intelligence would tutor millions of students simultaneously, I would have pictured something like the learning equivalent of a 24/7 patient teacher who never gets tired, never runs out of explanations, and somehow remembers exactly where each student got stuck. It sounded almost too good to be true. By late 2025, Khan Academy’s Khanmigo had logged over 50 million tutoring sessions since launching in 2023, and the numbers looked genuinely promising. That’s not a trivial achievement. But here’s what I’ve learned after watching this technology develop and listening to colleagues who’ve integrated it into their classrooms: the real story is far more complicated than either the hype or the skepticism would suggest.

Khanmigo 2.0 represents a meaningful shift in how we’re thinking about personalized learning at scale. The improvements in math completion rates among middle school students, which jumped 23% for those using the platform, tell us something real is happening. But as educators who care about how learning actually works, we need to understand what’s working, why some students gain dramatically more than others, and where this technology is quietly undermining something precious about the learning process itself.

Where the Science Actually Backs It Up

Let me start with what genuinely impresses me about Khanmigo’s approach. The platform has built in a principle from learning science that I’ve been teaching about for years: the Socratic method actually works better than direct instruction when the goal is retention and transfer. Research from Stanford Graduate School of Education published in 2025 found something striking: AI tutors that ask guiding questions outperform those that simply provide answers by 31% on retention tests given two weeks later. This is not a trivial difference. This is the kind of gap that suggests students are developing deeper understanding rather than just following along while someone explains something to them.

What makes this particularly effective is that Khanmigo has built this into its architecture deliberately. Rather than defaulting to “here’s the answer,” the system is designed to prompt students to think through problems themselves. When I watch students interact with it, I see them slowing down, reconsidering their work, and actually explaining their thinking out loud. That’s the move we want. That’s where learning happens.

The gains for English language learners have been especially noteworthy. A 41% improvement in reading comprehension scores after eight weeks is substantial enough that it deserves real attention. For students who are simultaneously acquiring English and engaging with academic content, having a patient tool that can adjust its language complexity and provide instant feedback, without the social anxiety of asking a teacher or classmate for the hundredth time, makes real sense. The Gates Foundation clearly agreed, committing $15 million in Q3 2025 specifically to expand access in Title I schools across twelve states. That kind of investment suggests policymakers are seeing genuine promise in these results.

The Struggle Problem Nobody’s Talking About Enough

But here’s where I need to be direct about something that’s been bothering me since I first heard about it. A January 2026 survey by the RAND Corporation found that 67% of teachers using AI tutoring tools reported their students became less likely to engage in productive struggle. Let that sit for a moment. Two-thirds of teachers noticed their students backing away from hard problems more quickly.

This matters because struggle is not the enemy of learning. Struggle is often the learning itself. When a student sits with a confusing concept for a while, makes a mistake, notices it, revises their thinking, and arrives at understanding through that process, something happens neurologically that’s different from being guided smoothly to the answer. The neural pathways that form through struggle tend to be stickier. More retrievable later. More transferable to new problems. We know this from cognitive science.

What I suspect is happening with some implementations of Khanmigo is subtle but important: students are learning that help is always just a click away. When the friction of not knowing something is reduced too far, when the tool makes it easier to get an answer than to keep thinking, students rationally decide to opt out of the cognitive work. I see this in my own classroom when I’m not intentional about when I deploy tools and when I deliberately hold space for productive confusion. The tool itself isn’t the culprit. The way we’re using it sometimes is.

What 2.0 Gets Right That Earlier Versions Missed

Khanmigo 2.0 has started addressing some of these concerns, at least in its design if not always in how schools implement it. The platform now includes better controls for educators to set parameters around how quickly hints are offered. You can configure it so students have to genuinely think before the tool jumps in. That’s the kind of design decision that shows the team understands the learning science here.

The efficacy data from Khan Academy Khanmigo Research and Efficacy also reveals something encouraging: the biggest gains aren’t appearing uniformly across all students. They’re concentrated where you’d predict the tool would be most useful. Students who previously had no access to individualized tutoring support are seeing the largest improvements. That’s exactly right. This technology fills a gap. For students in under-resourced schools, for students whose parents can’t afford tutoring, for students who struggle with asking for help in front of peers, Khanmigo is genuinely democratizing access to something valuable.

The Implementation Problem We Can’t Ignore

Here’s what concerns me most, though: Khanmigo itself might be excellent, but how schools are implementing it often isn’t. I’ve watched administrators roll out AI tutoring tools as a replacement for human connection rather than a supplement to it. I’ve seen it positioned as a cost-saving measure when budget cuts hit. I’ve heard about class sizes staying enormous because “now they have Khanmigo” to help students. That’s a betrayal of what the tool could be.

The research from Stanford PACE Center AI in Education Reports consistently shows that AI tutoring performs best when it’s part of a comprehensive learning environment, not a substitute for one. The students seeing 23% improvements in math completion rates, the ELL students seeing 41% gains in reading comprehension, I’d bet money that many of them are in settings where the AI tool is one resource among many, where they also have human teachers who care, who notice when they’re struggling productively versus actually stuck, who remember their names.

The question we should all be asking as educators, parents, and policymakers is this: Are we using Khanmigo to enhance human teaching or to replace it? The technology gets it right in its architecture and design. But we get it wrong sometimes in our choices about deployment. The tool isn’t the whole story. Our wisdom about how to use it is.

What This Means for Your Learning (Or Teaching)

If you’re a student considering using Khanmigo, here’s my honest take: it’s genuinely useful for targeted help on specific concepts, for getting unstuck when you’re truly stuck, for practicing once you already understand the basics. But don’t let it become a shortcut around thinking hard. The sessions that will actually change your brain are the ones where you struggle first, ask questions second, and understand why you were confused third.

If you’re an educator thinking about implementing this, the research is real and the potential is genuine. Use it intentionally. Set up your systems so students experience productive struggle before accessing hints. Use the time you save from grading routine practice problems to do what only humans can do: notice the specific confusion underneath a wrong answer, connect concepts to a student’s particular interests, celebrate the exact moment when understanding clicks. Build your classroom so the AI tool serves your teaching, not the other way around.

What’s your experience been with AI tutoring tools? Whether you’re using Khanmigo, considering it, or watching from the sidelines, I’d genuinely love to hear what you’re seeing work and what concerns you. The conversation about how we implement these tools matters as much as the tools themselves. Drop a comment below or reach out, I read every single one.

The Policy Wave Is Real, and It’s Moving Fast

Walk into your local school board meeting right now, and there’s a decent chance you’ll hear someone passionately advocating for phone-free school days. Something genuinely unprecedented is happening across American education. Since 2024, the number of states passing or introducing legislation that restricts smartphone use during the school day has nearly doubled—jumping from 8 states to 14 by January 2026. That’s not a slow cultural shift. That’s momentum.

What sparked this? A lot of credit goes to Jonathan Haidt’s 2024 book “The Anxious Generation,” which made a compelling case that constant device access might be harming young people’s mental health and development. The book resonated so deeply with educators and administrators that over 200 individual school districts adopted some form of screen restriction policy in 2025 alone. Then there’s Florida’s “Phone-Free Schools Act,” signed in 2023, which has become something of a model. By 2025, that policy showed measurable reductions in reported bullying incidents by 12 percent. When you see a number like that attached to something as pervasive and painful as bullying, it gets attention.

The Research Says Something Interesting (But Not What You Might Expect)

Here’s where I need to put on my researcher hat for a moment. The learning science picture is more complicated than the headlines suggest. Yes, there are some compelling findings. According to the UNESCO Global Education Monitor on Technology, countries that implemented full school smartphone bans saw a 14.23% improvement in baseline literacy scores among students who were already struggling. Let me be specific about what that means in practice: if you have a seventh grader reading at a fourth-grade level, a phone-free environment apparently helps them catch up faster. That matters.

And classroom teachers are reporting real changes. A recent Common Sense Media School Device Research survey found that 68% of teachers reported improved classroom focus after implementing partial device restrictions. Think about that percentage. It’s not 98%, which would be suspicious. But 68% of teachers noticing genuine improvement? That’s concrete, observable change that experienced professionals are documenting.

But here’s the honest part: the research on overall academic achievement and test scores is murkier. Some studies show modest gains. Others show minimal difference. The benefit appears strongest for students who were already struggling and for specific skills like reading and writing. The evidence for improved math performance, for instance, is considerably weaker.

What I’m Actually Seeing: A Worked Example

Let me ground this in something real. Two years ago, my school piloted a Wednesday phone-collection policy in just the freshman and sophomore hallways. On those days, students turned their phones into labeled pouches at the start of first period and retrieved them at lunch. I watched what happened, and I want to describe it exactly as I observed it, not through the lens of what I hoped would happen.

The most obvious change? Students asked more questions during class. Not necessarily better questions—just more of them. When you’re not half-focused on whether someone texted you, you notice when you don’t understand something. During those Wednesday units, my quiz scores didn’t dramatically improve, but my in-class engagement absolutely did. Students worked through problems with their peers more readily. The hallway conversations about the day’s lesson extended into lunch. These are not small things, but they’re also not the same as saying test scores went up by 15 percent.

The second observation was about anxiety, particularly around social comparison and FOMO. By mid-year, I noticed that on Wednesdays, the social dynamics shifted slightly. There was less visible tension around who was in which group chat. Friendships seemed less performative. By Thursday, once phones were back, that shifted again. I’m not a psychologist, and I’m not claiming this proves anything universal. But it suggested to me that the mental health component of phone-free time might be real, even if it’s harder to measure than reading comprehension.

The Confounding Variables Nobody Wants to Discuss

Here’s what makes this complicated. When a school implements a phone-free policy, they’re not just removing phones. They’re usually also creating a cultural shift. There’s more conversation about presence and focus. Teachers often adjust their own practices. There’s less passive screen time for students during transition periods, which means more actual movement and social interaction. Are improvements coming from the absence of phones, or from all these downstream changes in school culture? Both? We honestly can’t fully separate them yet.

There’s also a demographic reality that matters. Students from lower-income backgrounds often rely on phones as their primary internet access. Students with anxiety disorders might use their devices as a coping mechanism. English language learners sometimes use translation apps. When we remove devices entirely, we’re not removing them equally. The policy that feels freeing for one student might feel exclusionary for another.

What We Should Actually Do With This Information

So where does this leave us? The honest answer is that the research shows promise in specific areas, particularly for literacy in struggling students and for immediate classroom engagement, but not a wholesale academic transformation. The bullying reduction in Florida is real and meaningful. The teacher observations of improved focus are authentic. But this isn’t a silver bullet, and framing it that way does students a disservice.

The schools I’ve seen succeed with these policies treat them as one tool within a larger framework. They combine phone restrictions with explicit teaching about focus, with better classroom design, with teachers who are genuinely engaged. The phone-free time isn’t solving the problem by itself. It’s creating space for other things to happen.

What I want for every student in every classroom is what I’ve always wanted: the chance to think deeply, to ask questions without distraction, to experience genuine connection with their peers and their teachers. If a screen-free Friday creates that space, then it’s worth doing. But let’s also be honest about what the research actually shows versus what we’re hoping it will show. That honesty is what allows us to keep adjusting, keep improving, and keep meeting students where they are. What’s your experience been? I’d genuinely love to hear what you’re observing in your own schools and classrooms.

What’s Actually Happening in American Classrooms Right Now

Walk into most elementary school classrooms across America right now, and you’ll notice something shifting. Teachers are teaching phonics differently. Reading instruction looks less like guided discovery and more like deliberate, systematic instruction. This isn’t happening by accident. As of January 2026, forty-one states have passed legislation requiring structured literacy or Science of Reading-aligned instruction — a dramatic jump from just thirty-seven states in 2024. What started as a movement grounded in cognitive science has become policy mandate, and teachers are living in the space between research evidence and the reality of implementation.

The Science of Reading itself isn’t new. Decades of work from the National Institute for Child Health and Human Development showed us something remarkable: ninety-five percent of children can learn to read proficiently when taught with systematic phonics-based instruction. That’s not a suggestion or a best practice recommendation. That’s what the research actually says about human reading development. Yet for years, many teachers operated under different assumptions because that’s what their training emphasized and what their curriculum materials promoted. Now, suddenly, everyone is supposed to get on the same page — literally and figuratively.

The Evidence Works. Mississippi Proves It.

If you want to see what Science of Reading mandates can accomplish when implemented well, look at Mississippi. As the first state to fully implement Science of Reading reforms statewide, Mississippi has become the case study everyone references. Between 2019 and 2024, fourth-grade NAEP reading scores in Mississippi rose eight points. That might sound modest until you realize that Mississippi achieved the largest gain of any state during that period. Eight points means more kids reading at grade level. It means fewer struggling readers slipping further behind. It means the instruction is working.

The Mississippi story is compelling because it’s not theoretical. Those aren’t estimates or projections. Those are actual children reading better because their teachers received training in how reading actually works in the brain and implemented instruction accordingly. This is why the push for Science of Reading has so much momentum. The evidence isn’t questionable. The question is how to scale it successfully.

Here’s Where the System Is Struggling

This is where I need to be honest about what teachers are telling me and what the data actually shows. A 2025 RAND Corporation report found that fewer than forty percent of elementary teachers felt adequately trained to implement structured literacy approaches despite state mandates being in effect. Let that sink in. We’ve passed laws in most states. We’ve changed what we’re supposed to teach. But we haven’t effectively prepared the people doing the teaching.

Teachers aren’t saying this to complain. They’re saying it because they understand what’s at stake. Structured literacy isn’t just a different curriculum — it requires understanding phonemic awareness, orthography, morphology, and how these components connect to fluency and comprehension. It requires knowing how to diagnose reading difficulties and adjust instruction accordingly. It requires significant professional development, ongoing coaching, and curriculum materials aligned to these principles. Many teachers feel like they’re learning this alongside their students, which is anxiety-inducing when you’re responsible for someone’s foundational literacy skills.

The implementation gap exists partly because textbooks and curricular materials didn’t update overnight. For years, Lucy Calkins’ Units of Study curriculum was used in an estimated twenty-five percent of U.S. elementary schools. In 2023, following significant criticism from literacy researchers, Calkins’ program underwent substantial revision to incorporate explicit phonics instruction. That revision mattered because it represented acknowledgment that prior approaches were incomplete. But it also meant teachers using that curriculum had to completely rethink how they taught reading.

What Effective Implementation Actually Looks Like

Teachers who are successfully implementing Science of Reading talk about several key elements. They’re teaching phonics systematically and explicitly rather than assuming kids will absorb sound-symbol relationships through exposure. They’re assessing frequently to understand which students need which interventions. They’re connecting phonics to actual reading — not doing phonics in isolation but showing how these principles apply to real texts. And they’ve had genuine professional development, not just a one-day workshop.

The teachers who feel most successful aren’t necessarily the ones working in the most advantaged schools. Some are in under-resourced districts that finally received funding specifically for Science of Reading implementation. What they share is a combination of training, support systems, and often a principal or literacy coach who understands the research and can help translate it into daily practice. They’re also more likely to have access to curriculum materials actually designed around structured literacy principles rather than retrofitted materials.

If you want to understand what’s happening in your state or district, the Education Commission of the States Science of Reading policy tracker shows exactly which states have passed what legislation. More specifically, you can look at how your state defines Science of Reading requirements, what professional development is being offered, and whether materials adoption guidelines have changed. The variation in implementation quality largely depends on those specifics.

The Path Forward Requires More Than Policy

Here’s what keeps me thinking about this late at night: policy changes fast, but teaching culture and teacher preparedness change slowly. We’ve created the mandate structure, but we haven’t fully created the support structure. The evidence from reading science is clear. The policy environment is shifting rapidly. But the teachers in classrooms need more time, more training, more coaching, and more high-quality materials than most states have actually provided.

What needs to happen next is honest about the gaps. States need to fund professional development adequately. Teacher preparation programs need to restructure how they teach reading instruction. Curriculum companies need to continue developing materials that actually reflect structured literacy principles. And teachers need explicit permission to ask for support when they’re uncertain rather than feeling like they should have figured this out already.

The encouraging part is that the research is unambiguous. We know that systematic phonics-based instruction works. We’re seeing proof in classrooms like Mississippi’s. The challenge now is whether we can create the conditions that allow all teachers to teach reading this way consistently. That’s the real work, not the policy mandate itself, but the implementation that actually changes outcomes for kids. If you’re a parent, teacher, or administrator thinking about how Science of Reading might be affecting your school, I’d love to hear what you’re experiencing. What’s working in your classroom or district? Where are you seeing implementation challenges?

The Numbers Tell a Story We’ve Been Missing

Here’s what just happened, and I want you to sit with this for a moment. Community college enrollment grew by 4.7 percent for fall 2025. That’s the third year in a row we’ve seen genuine growth after the pandemic hammered these institutions. But here’s the part worth paying attention to: career and technical education programs are leading that charge. These aren’t general education courses filling up. These are welding certifications, nursing programs, cybersecurity bootcamps, and automotive technology tracks. Real, job-ready preparation.

At the same time, registered apprenticeship programs hit 770,000 active apprentices in 2025, up nine percent from the year before. We’re watching a genuine shift in how young people and their families are making educational decisions. And honestly, the conventional wisdom we’ve been feeding high school students for the past two decades hasn’t caught up to what the data is actually showing us.

The Economics Are Speaking Louder Than the Prestige Messaging

Let me show you what’s driving this change. The average student loan debt for someone with a bachelor’s degree now sits at $38,290. That’s not a small number. That’s a house down payment, a car, years of financial constraint right out of the gate. For a twenty-two-year-old trying to figure out if a four-year degree makes sense, that number has suddenly become impossible to ignore.

Now add this: according to recent research from the Georgetown University Center on Education and the Workforce, thirty percent of workers with associate degrees or occupational certificates are out-earning bachelor’s degree holders in their same age group. Thirty percent. Think about what that means for the conversation we’re having with students. We’ve been operating under an assumption that more education always equals more earnings potential. The evidence now shows that’s not universally true, and students are noticing.

When you combine the cost of a bachelor’s degree with data showing that a well-chosen associate degree or certificate can lead to better near-term earnings, you’ve fundamentally changed the cost-benefit analysis. Students aren’t being irrational when they choose community college. They’re being strategic.

Policy Is Finally Starting to Reflect What Works

There’s something else happening that tells me this shift is real and durable. California just passed Cal Grant Reform, effective January 2026, and it’s a genuine sea change. The state is now extending maximum financial aid eligibility to students in career education programs that last as little as sixteen weeks. Sixteen weeks. This is the first time in state history that short-term credentials have been treated the same way as traditional degree programs when it comes to state financial aid.

Why does this matter? Because policy is usually a lagging indicator. Policymakers respond to trends that are already happening, not the other way around. When a state as influential as California makes this move, it’s acknowledging something that the enrollment data, the apprenticeship numbers, and the earnings research have all been telling us: the pathway to economic security is no longer a single, standardized route.

This opens doors for students who might not have been able to afford a four-year degree but who can absolutely commit to and excel in a focused, shorter program. It also means that students who choose community college are making a decision that’s increasingly backed by both economic evidence and institutional support.

What This Means for How We Talk to Students

Here’s where the learning science piece becomes important. Neuroscience and educational psychology have taught us that people learn and decide better when we give them accurate mental models of how the world actually works. We’ve been giving students an outdated mental model. We’ve been telling them that a bachelor’s degree is the default, the safer choice, the one that leads to success. And we’ve been implying, sometimes directly, that community college is a fallback option.

The evidence suggests that’s not just inaccurate. It’s actually harmful to students’ decision-making. When you give someone incomplete or distorted information, they make worse choices. They might choose a four-year program they can’t afford, taking on debt that constrains their life for years. Or they might dismiss a community college path that would have been perfect for their interests, learning style, and financial situation.

What does better look like? It looks like sitting down with a high school student and saying something like this: “Let’s look at what you want to do. Let’s look at what it costs. Let’s look at what people are actually earning in that field with different types of credentials. Let’s talk about your learning style and whether you thrive in a large lecture hall or a smaller, more hands-on environment. Then let’s figure out which path gets you where you want to go most efficiently.”

Building a More Honest Conversation

I spent years as a traditional high school teacher, and I saw brilliant students choose paths that didn’t fit them because of outdated assumptions about what counted as a “real” education. I saw students with extraordinary practical talents get funneled into four-year universities where they floundered. Meanwhile, community colleges were preparing people for jobs that actually existed, with instructors who understood the industry inside and out.

The enrollment data tells us that more families are making smarter choices now. They’re consulting the National Student Clearinghouse Research Center enrollment reports, they’re looking at earnings data, they’re doing their homework. And they’re choosing community college and apprenticeships at record rates.

Your job as an educator, a counselor, or a parent is to make sure that choice comes from accurate information, not from desperation or financial constraint alone. Consult the Georgetown CEW education and earnings research. Look at regional job markets. Talk to people working in the field. Then have an honest conversation about what pathway actually makes sense for this particular student.

When we do that, something shifts. Students feel trusted. They feel like their unique situation matters more than a standardized prescription. They make choices they actually own, rather than choices they feel pressured into. They’re more likely to succeed, too, because they’re invested in a path that makes sense for them.

What conversations about college and career paths have you noticed shifting in your own circles? I’d love to hear what you’re seeing from the ground level, whether you’re a parent, a student, or an educator yourself.

The Numbers Don’t Lie, But They Don’t Tell the Whole Story Either

When the OECD released their PISA 2025 results in December, the headlines screamed the same story we’ve been hearing for years: we’re still recovering from the pandemic. Here’s what that actually means in concrete terms. Fifteen-year-old students across OECD nations scored an average of 472 in mathematics. That’s 7 points below where they scored in 2018, before everything shut down. Seven points might sound trivial until you realize that on standardized assessments, a gap that size typically represents about 2-3 months of lost learning. Multiply that across millions of students globally, and you’re looking at genuine disruption.

But here’s where it gets interesting, and where I think policymakers are missing the real conversation. Yes, we haven’t fully recovered. Yes, that matters. But the story changes completely when you look at who recovered fastest and why. This assessment cycle gives us something new to work with, something that finally points toward actual solutions instead of just documenting problems. You can explore the full OECD PISA 2025 Results Overview if you want the technical details, but let me walk you through what I think actually matters here.

Learning How to Learn: The Adaptive Learning Index Changes Everything

This is the part I genuinely wish every administrator in my district understood. PISA 2025 introduced something they call an adaptive learning index, which measures how well students self-regulate their own learning. It’s not about IQ or background or resources. It’s about whether a student knows how to notice when they’re confused, adjust their strategy, ask for help, or try a different approach. The findings are remarkable: students with high adaptive learning scores outperformed their peers by the equivalent of nearly two school years of progress. Two years. That’s not marginal improvement.

Think about what this actually means in a classroom. I have two students in front of me, same starting point, same curriculum, same amount of instruction. One of them has figured out how to be a strategic learner, someone who knows their own learning tendencies and adjusts accordingly. That student pulls in a resource when stuck. That student breaks difficult problems into smaller pieces. That student actually reads the feedback I give them. The other student, equally capable, just keeps doing the same thing and getting frustrated. The gap between them isn’t about intelligence. It’s about metacognition. It’s learnable. And yet most schools aren’t teaching it.

What Singapore Got Right (And Why It’s Not About Tutoring)

Singapore ranked first globally across all three PISA domains, with a mathematics mean score of 575. That’s nearly 100 points above the OECD average. I know what you’re thinking: Singapore must have incredible tutoring culture, massive test prep systems, harsh standards. Some of that is true, but the research shows the real story is different. Singapore invested heavily in teacher development and in creating structures where educators actually collaborate to understand why students struggle with specific concepts. They treat teaching as a profession that requires continuous learning, not a job where you figure it out and move on.

This matters because Singapore’s performance didn’t come from cracking down harder or testing more. It came from building systems where the adults in the room understood learning deeply enough to intervene effectively. When a student didn’t grasp fractions, Singapore teachers had protocols for understanding exactly where the misconception was and how to untangle it. That’s the difference between 575 and 472.

Teacher Collaboration as a Lever for Recovery: The Data Policymakers Keep Ignoring

Here’s the part that genuinely frustrates me about policy conversations. The OECD found that schools with structured teacher collaboration programs, where educators meet at least 2 hours weekly to analyze student data and share strategies, showed 18 percent stronger PISA score recovery rates between 2022 and 2025. Eighteen percent. That’s not a rounding error. That’s a meaningful, measurable difference in whether students actually recovered lost learning. And yet most policy energy goes into individual teacher evaluations, test prep, or curriculum overhauls, not into creating the conditions where teachers can actually work together to solve problems.

When I sit down with my colleagues and we look at last week’s quiz data, we’re not wasting time in meetings. We’re doing the work that actually changes outcomes. We’re noticing that 14 students got the same problem wrong. We’re asking each other what that tells us about how we taught that concept. We’re trying a different explanation next week and checking whether it helped. That’s not professional development theater. That’s actual learning science in action. And apparently, it’s one of the strongest levers we have for recovery. The U.S. National Center for Education Statistics – U.S. PISA Performance data showed American students scored 465 in mathematics on PISA 2025, below the OECD average for the second consecutive cycle, which prompted congressional hearings on federal funding priorities in early 2026. But those hearings probably didn’t focus on teacher collaboration time, which is free and available right now.

What Recovery Actually Looks Like

So if you’re a parent, teacher, or administrator reading this and asking what to actually do with this information, here’s where I land. Recovery isn’t a sprint. The pandemic didn’t just cost us 7 points in math scores; it fragmented our understanding of what it takes to help students learn well. But the PISA 2025 results give us a roadmap if we actually look at it. Students recover faster when they learn how to regulate their own learning. Teachers help that happen more effectively when they collaborate on understanding where students actually struggle. Schools that build these systems see real improvement, even when they start from behind.

If you work in education, ask yourself: Are we creating time and space for teachers to actually work together on data? Are we teaching students explicitly about their own learning processes? Or are we just hoping harder, testing more, and expecting different results? Recovery isn’t just about closing a 7-point gap. It’s about building the systems that make sustained learning possible. The data tells us exactly what works. The question is whether we’re listening.

The Report We Can’t Ignore Anymore

There’s a moment in every teacher’s year when you realize the data you’re holding isn’t abstract. It’s your students. Last week, the latest NAEP Nation’s Report Card landed, and if you haven’t sat down with it yet, you need to. The numbers are telling us something we can no longer address with band-aid solutions and good intentions.

Fourth-grade reading proficiency is still four points below where we were in 2019. Four points might sound small until you translate it into faces: roughly 240,000 students reading below grade level. And here’s what makes my stomach turn a little — that gap is widest in rural, low-income districts. The kids who need the most support are getting the least. That’s not an accident. That’s a design problem, and we can fix design problems if we’re intentional about it.

Why This Gap Costs More Than We Think

Let me ground this in something more concrete than percentage points. McKinsey’s latest analysis estimated that unaddressed learning loss could cost affected students up to $49,000 in lifetime earnings if the gap isn’t closed by age 18. Let that number sit for a second. That’s not just about test scores. That’s about trajectories. That’s about whether a kid from a rural district gets to make certain choices about their future or whether those choices get made for them.

The research shows us that learning loss from this period isn’t evenly distributed. It clusters. It compounds. If we don’t interrupt it deliberately right now, during semester planning, we’re essentially designing it into the system for the next five years. This is the year where your curriculum choices actually matter more than most years.

The Evidence for What Actually Works

Here’s what I love about teaching in 2025: we have better data than ever about what moves the needle. A comprehensive meta-analysis from the University of Chicago Education Lab just released findings on tutoring interventions, and they’re specific enough to actually guide your planning. High-dosage tutoring, meaning three or more sessions per week with a consistent tutor, showed a 0.19 standard deviation improvement in math outcomes. That’s real. That’s reproducible. That’s the kind of specific mechanism you can actually build into your semester.

The problem? Only 23% of schools that received federal pandemic relief funds used that money for evidence-based tutoring programs before the September 2024 deadline, according to the Education Recovery Scorecard Harvard. Most schools spent relief money on things that didn’t move the needle. That’s not a judgment on those schools — it’s an observation about how hard it is to align resources with evidence when you’re drowning in the day-to-day. But it also means there’s space for you to do this differently in your own classroom.

What Tennessee Got Right (And What We Can Learn)

I want to highlight Tennessee’s literacy initiative because it’s the closest thing we have to proof that statewide strategy actually works. Their third-grade reading proficiency jumped 8 percentage points in 2025. Eight points. In one state. That’s the largest single-state gain in this NAEP cycle. It didn’t happen by accident, and it didn’t happen with one intervention. It happened because they designed their entire system around literacy recovery.

Tennessee didn’t just add tutoring. They redesigned their curriculum sequence to target specific skill gaps early. They trained teachers intensively. They built accountability into how they measured progress. Most importantly, they didn’t treat literacy recovery as a special program — they made it structural. That’s the difference between a well-meaning initiative and one that actually moves the needle.

How to Actually Use This in Your Semester Plan

So what does this mean for you sitting at your desk right now, planning your semester? Here’s what I’d do. First, be honest about where your students actually are, not where your curriculum guide says they should be. If you teach fourth grade, assume your readers are probably a few months behind pre-2019 benchmarks. That’s your baseline. Second, identify your high-dosage intervention opportunity. Can you create a consistent tutoring structure, even informally? Can you partner students strategically? Can you build 15-minute focused sessions into your weekly schedule for the students who need them most? It doesn’t have to be a formal program. It has to be consistent and targeted.

Third, sequence your curriculum with recovery in mind. This isn’t about covering less material — it’s about designing a better sequence. Spend more time on foundational skills before moving forward. Build in more review cycles than you normally would. The goal is to create conditions where students can actually catch up, not just feel rushed through material they don’t understand yet. Use formative assessment seriously. Every two weeks, check where students actually are. Adjust. Don’t just push forward and hope.

Fourth, protect consistency above novelty. The research is clear: students recovering from learning gaps need the same tutor, the same approach, the same classroom environment. Consistency beats flashiness. If you’re planning new curriculum, make sure you’re not sacrificing that consistency for variety.

The System We’re Building Right Now

Here’s what I keep thinking about. Every curriculum choice you make this semester either perpetuates the gap or interrupts it. There’s no neutral. If you design your semester as if all your students had uninterrupted education pre-2020, you’re passively accepting the learning loss. If you design it with recovery built in, you’re actively intervening. The data shows you which students need this most. Rural kids. Low-income kids. The students who probably need you the most and get you the least.

This report isn’t meant to depress you. You have better evidence than ever about what works. You have access to the outcomes. You have a semester to plan. And you have 150 students (or however many are in your classes) who are counting on you to design something that actually gets them where they need to be, not just where the calendar expects them to be.

I’d love to hear how you’re thinking about recovery in your own classroom planning. What patterns are you seeing? What’s one specific intervention you’re building into your semester? Drop a comment or reach out — this is the kind of work that gets better when we think through it together.