Why This All Matters (For People Who Don’t Pay Attention to Higher Education)

My friend Kevin asked me, in good faith, why any of this should matter to him. I owe him a real answer.

An old friend, kevin jones, dropped a comment on my last piece that's stayed with me. The question was kind and straightforward: "I do not pay attention to education. Can you explain why this matters and what the harm is, and to what?"

I mean, hey, Kevin's right to ask. Most higher-ed news coverage assumes the reader already cares, and if you don't, the whole thing reads like insider baseball academic shenanigans (which, well, it is some most of the times). This post started with my reply to him, which is here if you're curious.

That was the short version. This is short-medium (for a SCBBQ post anyway) version with some more thought put into it: why what’s happening to American higher education affects several things you probably do pay attention to, even if you’ve never thought about higher ed as such.

---

Higher education is going through a pretty substantial structural reshuffling, and it likely won’t be finished for another decade. But the reshuffling will matter for a lot of people who don’t track the sector, and here’s why.

The most immediate harm is to students and communities at the “bottom” of the institutional hierarchy. About 40% of American four-year colleges are in a fiscally precarious zone. A recent Huron Consulting projection puts 442 of the country’s 1,700 private nonprofit institutions at risk of closure or forced merger within ten years, with 120-plus at highest risk. Sterling College in Vermont is closing this semester at 40 students, down from a peak of 120. Bloomfield College in New Jersey was absorbed by Montclair State. Birmingham-Southern in Alabama closed in 2024. Labouré College in Massachusetts is ceasing operations this August. When colleges at the bottom of the market close, their students lose continuity, some don’t finish, and the towns built around those colleges lose their largest employer, their housing demand, their cultural infrastructure. This is already happening across the Midwest and Northeast, and it accelerates over the next decade.

(This is the High Stress category in the university mapping of the structural divide I’ve been doing here, which if you’ve been a reader, I need not send you a link to (“Dude, Kyle, seriously, we know about the Mapping Research, my man.”) again. There is a link at the very bottom of this post if you’ve not seen it.)

The second harm is to research. American research universities have led the world in medical, scientific, and technological research for seventy years. The mRNA vaccine technology, the internet, penicillin, much of modern AI: these came out of that research compact. Now the new endowment tax (up from 1.4% to 4–8%, depending on institution), plus federal research cuts, plus state-level tenure restructuring, is eroding the pipeline. Yale just cut PhD enrollment 13% in arts and humanities, 5% in sciences. MIT is closing library service desks and freezing raises above $85K. Those cuts compound. Fewer graduate students now means fewer future faculty in a decade means fewer researchers in fifteen years means fewer discoveries. The effects show up slowly, but they compound, and the country slowly loses its research edge to competitors who are investing rather than squeezing.

The third harm is to the labor market. When employers can’t trust what a degree means, they narrow who they hire and people without elite credentials get screened out in new ways. As of January, a majority of Indeed job postings don’t require a college degree. But Lumina research finds that even employers who drop formal requirements still prefer degree holders for 48% of roles. The labor market is adjusting around the sector’s unreliability, but not uniformly in ways that help most workers. Skills-based hiring sounds democratic in press releases; in practice, it often reverts to informal networks and employer-specific certifications that favor people already inside the door.

The fourth harm is to the top-end pathway of social mobility specifically, and here’s where I should ask you for a bit of the nuanced thinking. The BA was the main ticket into the American middle class for seventy years. A bifurcated credential might actually expand access at the mass bottom-to-middle level. A cheap, fast, accredited BA can let working adults credential up without debt or quitting their job. That’s plausibly more mobility at the mass scale. But Raj Chetty’s Opportunity Insights research shows that a small number of selective institutions play an outsized role in pulling working-class students into the top tier. If the credential bifurcates, the kid who stacks Sophia credits gets a cheaper BA and a discount labor market. The kid who goes to Stanford gets the network and the moat, and those tighten at the same time. The bottom-to-top pathway narrows even as the bottom-to-middle pathway may widen. Whether that’s net positive or negative for mobility depends on which you weight, and honest people will land differently on it.

The fifth harm, contested but directionally clear, is to public knowledge and trust. Universities historically weighed claims, trained citizens to reason, and produced the public intellectuals who shaped debate. Gallup’s trust data has confidence in higher education at 42%, up from a historic low of 36 but well below the 57% it held in 2015, with a 40-point partisan gap (66% Democrats, 26% Republicans). Yale’s own Committee on Trust in Higher Education dropped a 58-page report earlier this month with a Yale president explicitly acknowledging that universities helped erode the trust they now lack. If that function degrades, the whole information environment gets worse, at a moment when we need it to get better.

The honest summary: the institutional scaffolding of American upward mobility, American research capacity, and American regional economic stability has been weakening in parallel, and most of us (including me, until recently) have not taken seriously what that looks like on the other side.

---

That said, there is real reason for some cautious optimism for actual education if you look at what’s being built outside the traditional sector.

AI tutoring is actually good now, and getting better. Math Academy, built by Justin Skycak, publishes (self-reported, not yet independently validated, so he could be “talking his book” but still…) that students learning STEM material advance 2–3x faster than in traditional instruction, using hierarchical knowledge graphs with mastery-based progression and spaced repetition. The selection caveat matters: students who choose Math Academy are already motivated learners, so some of that speed is the selected population, not just the product. But the underlying approach is plausible and scalable. Khan Academy deployed Khanmigo as a personalized AI tutor at scale for students and teachers. Anthropic’s Claude and OpenAI’s tools increasingly function as on-demand tutors for motivated learners. Benjamin Bloom’s famous “2-sigma problem” from 1984, that one-on-one tutoring produces outcomes two standard deviations better than classroom instruction, was unreachable at scale for forty years because of cost. AI makes the cost problem tractable. Whether it solves the quality problem at scale is still being tested. If it does, learning itself becomes dramatically more effective for whoever has access.

One honest caveat on AI tutoring. It works for motivated learners who engage with material as students. For users who treat AI as an answer machine rather than a tutor, which is a large share of current student use, AI probably hurts retention and critical thinking rather than helping. MIT Media Lab research (Kosmyna et al., 2025) found reduced cognitive engagement and weaker memory formation in students who offloaded essay-writing to ChatGPT compared to students using search or brain-only approaches. A majority of students themselves now report believing heavy AI use is harming their critical thinking, per recent Lumina-Gallup polling. The "cognitive offloading" problem is real. Whether AI makes learners smarter or lazier depends on how it's integrated into practice and how students are socialized to use it, not on the technology alone.

Employer-built credentialing is scaling, but quite unevenly (so far). Google Career Certificates have partnership agreements with hundreds of employers for specific technical roles. Whether hiring managers actually weight them equivalently to BAs in practice is a separate question, and early evidence suggests employers still discount them. AWS, Microsoft, Cisco, and Salesforce run certification systems that hiring managers do read. Some employers are building parallel educational infrastructure outright: Starbucks’s College Achievement Plan with Arizona State has served tens of thousands of baristas since 2014; Walmart’s Live Better U offers debt-free college to its workforce. These programs serve existing employees rather than providing entry-points for outsiders, which is a real limit worth naming. IBM, Siemens, and JPMorgan all run white-collar apprenticeship programs. The pattern is that large employers are producing the workers they can use directly, in parallel to the sector.

Apprenticeship is rebuilding as an alternative pathway. Registered apprenticeships more than doubled between 2014 and 2024, with roughly 680,000 active participants. That’s still small relative to the roughly 19 million in undergraduate enrollment (about 3.5%). Real and growing, at a scale that matters for specific industries more than for the general labor market. About 350 institutions now offer some form of degree apprenticeship. White-collar apprenticeships in cybersecurity, healthcare IT, and teaching are expanding. Both the federal government and major employers are investing. This is the pathway that most resembles Germany’s dual-track system, which took decades of institutional infrastructure to build there. We should expect it to take a decade or more to build at similar scale here (but I’d argue that’s taking wayyyyy too long).

K-12 acceleration models are changing what students will arrive with. Schools using the Alpha School model, including its franchised variants like Texas Sports Academy, have students complete core academics in 2–3 hours of AI-assisted morning work, with afternoons for projects and specialized pursuits. Alpha School’s internal reports have 5th graders completing 8th-grade material. If those patterns hold, and it’s still early, the students arriving at post-secondary in the next decade will have different expectations about what learning should cost in time.

Community-based learning and trade pathways are reasserting themselves too. Community colleges still serve more Americans than any other sector of higher education and are better positioned than most observers recognize to be the on-ramp to skills-based employment. Trade unions are running their own apprenticeship pipelines with earnings outcomes that can beat many bachelor’s degrees. The infrastructure for learning-plus-credentialing-without-traditional-college has been quietly building for a decade, mostly outside the attention of people who cover higher ed.

If the traditional sector bifurcates and thins, the pieces that matter most (real learning, trusted credentialing, labor-market signaling) might reassemble outside the university in certain sectors. That’s a conditional with real weight. Apprenticeship plus employer certification plus AI-enabled tutoring plus skills-based hiring all have to mature together for the alternative to work. Any one of them stalling breaks the system. And credential mechanics alone don’t deliver mobility. Kids with wealthy networks keep getting internships, family connections, and the cushion to take unpaid opportunities, whatever the credentialing infrastructure looks like. The 2015-era coding bootcamp boom tried something close to this and mostly failed at scale, with tech employers returning to CS degrees for serious hiring. But if enough of the pieces do cohere, the reassembled system could deliver more mobility than the current degree-centric one. That would be a good outcome, even if it’s painful for the institutions that don’t adapt.

Of course there’s no guarantees on any of this, but the harms are in motion. So is the rebuilding. AI is what makes the degree-hacking and AI-cheating stories possible, and also what makes tutoring-at-scale plausible. Same technology doing both things at once. Which wins out depends on who builds what, and who pays attention.

Which is why I wrote the long version of this argument in the first place, with the full data and structural analysis behind it. Read that if you want the whole thing.

But if you only take one thing from this short version, it’s this: the restructuring of American higher ed is real and it will affect the labor market your kids enter, the research that produces the next generation of medicines and technologies, the economic health of towns you pass through on the interstate, and the trust infrastructure of American public knowledge. The harms are real. The opportunities are also real. The next decade is when a lot of this gets built or doesn’t, and who pays attention now shapes what gets built.

Sacred Cow BBQ covers higher ed, politics, institutions, and the trust questions underneath. If this was useful, subscribe for more. The interactive university map is available if you want to see where specific institutions land.