Technology in the Classroom: What Actually Has to Change
Written By: Ryan Morrison
Listen On: YouTube | Spotify | Apple Podcasts
2026 Asset Allocation Report
See how 230+ HNW investors with an average net worth of $17M are allocating across public equities, private markets and alternative assets.
Most schools are responding to pressure about technology by adding more of it. Michael Horn's argument is that this is exactly backward — not because technology is bad for learning, but because layering tools over a broken instructional model predictably makes that model worse, not better. The conversation that follows moves from how parents manage phones at home through how schools should actually be using AI, to what the job market shift means for kids who are about to graduate, and whether elite colleges still make sense in 2026.
Horn co-founded the Clayton Christensen Institute and has spent two decades studying how education systems change. He teaches at Harvard, co-hosts the Future You podcast with bestselling author Jeff Selingo, and recently published Job Moves, a research-based guide to navigating career transitions. He joined Tad Fallows, Matt Shechtman and Sriram Gollapalli on Navigating Wealth — all four are parents with school-age kids navigating these decisions in real time.
TL;DR
Managing screens at home is less about time limits and more about developmental stage, social context and what the device is replacing
Technology doesn't improve education when layered on a broken instructional model — the model has to change first
Commercial EdTech products set mastery thresholds at 70-80% rather than 90%+, which explains why EdTech hasn't delivered the gains its advocates promised
The meaningful line on AI in schools is between tools that sharpen thinking and tools that offload it entirely
CS majors currently face roughly double the unemployment rate of art history majors — the "learn to code" safety narrative has inverted faster than most expected
55-85% of jobs were filled by network before AI flooded the application pipeline; that percentage is likely to increase
If admitted to a highly selective school, the research still supports attending — but the value has always been network and credentialing, not curriculum
Students who go to college primarily to meet others' expectations drop out or transfer at a 74% rate, according to Horn's research
Screen Time, Phones and the Outcast Problem
Managing screens at home is less about the device and more about the social and developmental context of the child.
Horn frames it as a staged evolution. For young kids, minimal and interactive — FaceTime with a grandparent, a parent sitting alongside. Passive consumption as a babysitting mechanism is what he's trying to avoid. As kids get older, the framework shifts toward graduated freedom, clear household guidelines and explicit conversations about what healthy habits look like in your family.
His family has adopted a no-smartphones-before-8th-grade posture, influenced in part by the arguments Jonathan Haidt and others have made about the addictive dynamics of social media. Horn is careful to note that his concern is less with phones per se and more with the social media access those phones provide before kids have the prefrontal cortex development to handle the dopamine triggers involved. "We struggle as individuals to do that stuff, let alone young kids and teenagers," he said.
The practical rules his family has settled on: devices out of the bedroom at night, no social media apps until developmental readiness, and parents modeling the behavior they're asking for. That last one came up directly in the conversation when Tad shared that his 10-year-old daughter caught his wife checking texts at a school concert and told her to put the phone away.
"It's great when they call us on it," Horn said. "Owning that as a parent is actually a cool opportunity to have the kinds of conversations without it being about your kid."
The harder question — one Tad and Sriram both raised from their own experience — is what to do when withholding a device starts to create an outcast situation. When your kid is the only person in the orchestra group text without a watch, or the only seventh grader who has to walk to the front office to make plans, the calculus changes. Horn acknowledged this directly.
"You've got to find the right solution that doesn't make your kid feel like an outcast and is pro-social behavior. That's the ultimate purpose — keeping the outcome in mind." The guardrail he keeps returning to is the difference between technology that becomes anti-social, sending a child down a rabbit hole of isolated consumption, and technology that supports connection. The former is what he's guarding against.
The Social Media Problem Is Different From the Screen Problem
Horn draws a distinction that most parent conversations about phones collapse together. The device itself is not the primary risk. The social media ecosystem — the addictive product design, the dopamine feedback loops, the group dynamics that emerge when 12-year-olds are interacting through text rather than face to face — is the specific thing he's trying to delay.
He surfaced an observation that came up in his own household: his kids came home describing a text exchange between two close friends where one had written something hurtful. "Do you think she would have said that if they were face to face? No." The medium lowers inhibition in both directions — it makes cruelty easier and accountability less visible.
His summary of the broader dynamic: "We've let kids into the online world with very little supervision, and we've corrected the other way on the offline world with lots of supervision. Moderation is a virtue in most places."
Beyond Wealth Newsletter
How HNW founders and executives navigate the questions wealth creates — grounded in peer data and Long Angle community discussions. Free, delivered every Thursday.
Why Layering Technology on a Broken Model Fails
Technology improves education when it enables a fundamentally different instructional model — not when it is added to the traditional classroom without changing how teaching works.
The traditional model Horn describes is built around 20 to 30 kids, a teacher at the front of the room, everyone moving at roughly the same pace through the same activities regardless of mastery level, background knowledge or interest. If you introduce phones or devices into that model to run a quiz or pull up a resource, you haven't changed anything meaningful. You've digitized a model that was already flawed.
COVID demonstrated the worst version of this. "Let's take the traditional model, put it online and Zoom with young kids and somehow expect this to work out well. And by the way, half of them might be in person while the other half are virtual and we're asking a teacher to manage both. It's like the dumbest thing I could possibly imagine trying to do."
The version that actually works looks different. Thirty minutes a day on a constrained, adaptive learning app — targeted to where each student actually is on a specific skill, feeding data back to the teacher — followed by the teacher using that data to run small-group instruction with kids who are struggling with the same thing. The technology is doing something the teacher couldn't do alone at scale. The model has changed. That's the distinction.
What's driving schools to default to the layering approach instead? Horn points to a real problem. Roughly 22% of students were chronically absent in 2024-25, remaining well above the pre-pandemic baseline of 15% — and chronic absenteeism is consistently ranked among district leaders' top challenges. Teachers are looking for tools to reengage students who have partly disengaged from school as an institution. Technology at the margins can help with that. But it's not the answer.
The Alpha School Model — What's Actually New
The Alpha School model — roughly two hours of intensive AI-enabled academic instruction followed by mentor-based experiential learning — generates significant discussion among Long Angle members, and Horn knows the founders. His assessment is measured.
The core idea is not new. Homeschoolers have demonstrated for years that right-sized academic content, delivered without the transition time and pacing constraints of a traditional school day, can be compressed significantly. Acton Academy in Austin and Summit Public Schools in California have built similar models — compressed academic time followed by substantial project-based blocks — for many years.
What's specific to Alpha School's implementation is an observation about commercial EdTech products that Horn found genuinely interesting: the mastery thresholds in widely used platforms like Khan Academy were set at 70-80% rather than 90%+. "Theoretically, you shouldn't be able to move on from single-digit addition until you actually demonstrate mastery of single-digit addition. And we weren't necessarily seeing those gains." That's a specific, non-obvious mechanism behind a broadly discussed phenomenon — EdTech products were designed for a market that requires ease of use and teacher adoption, not maximum learning outcomes, and those incentives are in tension.
AI in Schools — Where the Line Actually Falls
AI is eliminating the entry-level positions that have historically provided the on-ramp to professional careers, and schools are not built to respond to that.
The most specific data point in the conversation: CS majors currently have roughly double the unemployment rate of art history majors. Horn surfaced this as a direct challenge to the "learn to code" narrative that many HNW parents have absorbed as conventional wisdom for their kids. The implication is not that technical skills don't matter — they do. But technical skills divorced from the capacity to ask whether an AI-produced outcome is actually what we want, or to manage the human relationships that close deals and build organizations, may be less durable than assumed.
His practical advice for young people entering the workforce now comes in three parts. First, prioritize real work experience and demonstrated building over academic enrichment programs. "How do you use AI to be a creator, to build something of value and meaning? Those are the sorts of things I would be asking." Second, learn to make tradeoffs in the job market rather than waiting for a job that checks every box. "Every job has elements of suck in it. The question is which elements you're signing up for." Third — and Horn was direct about this — optimize for mentorship access early in a career, not status or title.
The networking dynamic he described is one that Long Angle members who do hiring will recognize immediately. Somewhere between 55-85% of jobs were filled by network even before AI automated the resume-and-application process. A hiring manager who needs someone doesn't look at applications — they call someone they trust and ask for recommendations. That process is accelerating, not going away.
The structural response Horn advocates for is permeability between schools and the workforce. Teen summer employment peaked above 50% in the late 1990s before declining sharply — and the gap in basic work competencies that this created is real. His prescription: start in middle school with career exposure, not to build occupational identity (that job may not exist in 10 years) but to begin understanding what kinds of work produce energy and what kinds drain it. Move to learning through real work in high school — community-embedded projects, internships, youth apprenticeships — so that by the time a student reaches college or the workforce, they have at least some reps. Pew Research Center
The CS Unemployment Inversion — What It Means
Horn's take on the inversion is not that STEM is wrong. It's that STEM married to humanities becomes more valuable, not less, as AI handles more of the technical execution. "How do we manage trust? How do we build relationships? Like, yes, AI could talk to AI and ink deals, but more likely I think is that people will be doing a lot of that." The care economy also appeared as an underappreciated growth area — jobs that require human presence and human warmth in ways that AI companionship cannot fully replicate.
One structural note that bears on the workforce training conversation: 65% of college students still attend within 50-75 miles of where they live. The education and workforce systems are still local in most cases, which means the most tractable version of Horn's school-workforce integration model is rooted in community — local employers, local governments, local nonprofits — not national tech companies.
Most of the conversations parents at this wealth stage want to have aren't happening in mainstream media.
They're happening between people who are navigating the same decisions — school choice, phone timing, college fit, how to position kids for a workforce that's changing faster than the conventional advice accounts for. Long Angle is where those conversations happen, candidly, with peers who have already been there.
College Selection — What the Evidence Actually Shows
If admitted to a highly selective school, the research still supports attending — but the value has never been the curriculum.
Horn is direct about this. The network and credentialing function of elite institutions is the point. Self-selection of a high-quality peer cohort, the screening signal that comes with admission, the marriage and business partnership patterns that flow from shared time in the same institution — these are durable. Tad put it plainly: "You could have gone to a public library and checked out a textbook and read all about calculus. I don't think it was ever really the point of the subject matter." Horn agreed: "Less than 5% of us are autodidact learners."
For students who do get in, Jeff Selingo's research — Selingo co-hosts the Future You podcast with Horn — suggests elite schools offer roughly double the probability of landing in the top tier of economic outcomes compared to similarly priced but less selective alternatives. Horn finds that a compelling data point: "That seems like pretty good odds."
The counterintuitive finding he raised on school-fit research: STEM majors at elite schools disproportionately abandon STEM because the competition is too intense. Being a big fish in a smaller pond is documented to produce better outcomes in some cases, particularly for students who would have been strong in their field at a less competitive institution but get ground down at Stanford or MIT.
The most actionable piece of the college conversation for parents was the dropout data from Horn's own research. Students who go to college primarily to fulfill others' expectations — without a genuine internal reason for being there — drop out or transfer at a 74% rate. "When going got tough, if they didn't have a deeper why, they were out." The implication for parents of college-age kids: before optimizing for prestige, make sure there's a real answer to why this child is going to this school at this time.
For students who genuinely aren't ready or aren't excited, Horn endorses a gap year — with structure. Real work, a service component, some form of active learning. Not aimless travel, though he'll fund the travel if it's part of a broader experience. The purpose is to light a spark rather than fill time.
Long Angle members across parenting ages are working through versions of this question — from what phones to give a 12-year-old to whether a gap year makes sense for a kid who got into a good school but doesn't seem ready. For parents comparing notes on these decisions with peers who are navigating the same complexity, see also how to teach kids about money when you can afford to give them everything.
Frequently Asked Questions
What is the best way to manage screen time for kids?
Manage screens based on developmental stage and social context rather than a fixed time limit. Young children benefit most from interactive, parent-involved screen use; passive consumption used as a substitute for engagement is the primary risk. As kids get older, the framework shifts toward graduated freedom, clear household rules and keeping devices out of bedrooms at night. The social media problem is distinct from the screen-time problem — most experts advocating for delayed phone access are specifically concerned about social media access before prefrontal cortex development can handle the dopamine feedback dynamics involved.
Does technology in the classroom actually improve learning?
Technology improves learning when it enables a fundamentally different instructional model — specifically, adaptive tools that target individual student skill gaps and feed usable data back to teachers for small-group instruction. Technology does not improve learning when it is layered on the traditional classroom model without changing how instruction works. Most schools default to the latter, which is why the EdTech gains promised over the past decade largely haven't materialized.
Should students use AI tools for homework and assignments?
The meaningful line is between AI that sharpens thinking and AI that replaces the cognitive work the assignment was designed to produce. Purpose-built educational tools that target specific skills are defensible. Using an LLM to generate a draft and submit it without engagement is not. The more productive framing is the sharpening use case: using AI to surface counterarguments, identify gaps in your own reasoning, or understand what a stronger version of your argument would look like. Schools that have only addressed this as a cheating problem have not yet asked the harder question of what the assignment itself should look like.
Is a computer science degree still worth pursuing?
CS remains valuable, but CS majors currently face roughly double the unemployment rate of art history majors — a specific inversion of the prevailing narrative. The combination of technical skills with humanities or domain expertise is increasingly what employers are looking for as AI handles more of the execution layer. Technical skills divorced from communication, judgment and relationship-building capacity may be less durable than a decade ago.
Is an elite college still worth attending in 2026?
If admitted to a highly selective school, the research supports attending. The network, credentialing and peer cohort effects are real and durable — they have always been the primary product of these institutions, not the curriculum. Jeff Selingo's research suggests elite schools roughly double the probability of landing in the top tier of economic outcomes versus similarly priced alternatives. The counterintuitive caveat: STEM students at elite schools disproportionately abandon STEM due to competitive intensity, and the big-fish-small-pond research suggests this is a real tradeoff worth weighing for the right student.
What should young adults prioritize when entering a job market shaped by AI?
Real demonstrated work and building experience outweigh academic credentials right now. Horn's three priorities: get real work experience earlier rather than adding more enrichment programs; learn to make job market tradeoffs rather than waiting for the perfect role; optimize for mentorship access over title or status early in a career. Networking is not optional context — 55-85% of jobs are filled through network, and that percentage is likely increasing as AI floods the formal application process.
What is the "wait until 8th" phone approach?
A posture popularized by Jonathan Haidt and others that delays smartphone access until 8th grade as a strategy for protecting adolescent development from social media's addictive dynamics. Horn's family has adopted a version of this approach. He is careful to note that the core concern is social media access, not phones as devices, and that the outcast-risk tradeoff is real — families need to weigh the social cost to their specific child in their specific community against the developmental rationale for waiting.
Final Thoughts
The debate about technology in education tends to stay at the surface level — more screens or fewer, ban phones or allow them, AI is cheating or AI is the future. Horn's argument cuts underneath all of that. The question is not which tools schools are using. It is whether the model of learning itself has changed in ways that make those tools meaningful.
The parents best positioned for the next decade are not necessarily the ones who chose the most technology-forward school, or the ones who kept phones away the longest, or the ones who pushed hardest for an elite college. They're the ones who understood what durable skills their kids were actually building — the social dynamics, the teamwork, the capacity to tolerate boredom and navigate real-world complexity — and made intentional choices to protect those things, in whatever school and at whatever wealth level those choices had to be made.
The decisions Horn describes are exactly the ones Long Angle members compare notes on every week.
School selection, technology access, college fit, workforce positioning — these are high-stakes calls without obvious right answers. Long Angle gives you a vetted peer group that's already navigating the same complexity, in a confidential environment with no pitches.
Resources Mentioned
Michael Horn’s Substack – The Future of Education newsletter.
MichaelBHorn.com – Michael’s writing, books, and research.
Michael Horn on LinkedIn – Where he shares ongoing analysis of education and workforce trends.
