The great unbundling of higher education
Universities are broken. A hacker hotel in Helsinki shows what might come next.
I heard it suggested recently that the current moment is reminiscent of the explosion of knowledge in the second half of the 15th century.
After Gutenberg’s invention of the moveable type printing press around 1450, and its ensuing wildfire spread across Europe, there came a mass internalisation of thought and learning. All of a sudden lay people could access information in the vernacular - not just religious texts but secular works on anything from science to romantic literature to history.
With this spread of literacy and learning, the hegemony of information and ideas fell out of the church’s grasp - the most notable result of which was of course Martin Luther, the grumpy pot bellied Saxon who took on the might of the Roman Catholic Industrial Complex and won. Alongside this fragmentation of Christianity came the spread of scientific and philosophical ideas that powered the Renaissance - a novel technology that unlocked widespread access to information that was previously gate-kept or fragmented and in turn drove forward human progress.
Sound familiar? What if 2026 is like this again but for a monopoly over learning? This is my attempt to grapple with whether large models will do for higher education what the printing press did for the church’s dominance of knowledge.
Big ideas, bigger debts
I read history at a UK university in the mid 2010s, emerging from my degree with a wonderful handful of gobbets about the past (a worrying amount of which were about Martin Luther’s love of beer), a network of smart, ambitious friends, a solid stamp on my CV and +£40k of debt.
Even at the time the debt felt hefty – tracking well above inflation, but manageable as it was salary linked. And this was 2016 - the days when a graduate with a half decent degree could walk into a job.
And importantly – it was before LLMs. The internet was already an astonishing resource, but it had to be traversed, sifted, waded through. It rewarded the incredibly diligent, the patient and the brilliant, but was nowhere near the autodidact’s paradise large models have turned it into.
2026 doesn’t look much like 2016. Graduate jobs continue to crater across the developed world, leaving young people sending bulk applications into a void of soulless AI interviews and constant rejection. It’s a hard sell to take on three years of debt for an uncertain outcome and a potentially redundant skill on the back end.
While on the flip side we’ve stumbled upon one of the most powerful tools man has ever invented. Large models could well be the next enabler humanity has discovered in the long line from fire, through agriculture, minerals extraction, to moveable type, the steam engine, electricity and the transistor.
As I’ve written about at length, a smart young person with a Claude account and a few hundred pounds in credits can not only up-skill themselves on essentially any topic, but then act upon that up-skilling – building anything from PCBs to consumer electronics to stone carving robots to software products.
And this is all as we witness the decimation of British universities. Their finances are crumbling, with 45% (124) of UK universities projected to face a deficit in 2025-26 (and 1/6 having less than 30 days of cash on hand). As a result, these institutions shed 13,300 jobs through severance in 2024-25 alone, with more than 90 announcing staff redundancies, course closures or other restructuring, from Nottingham to Essex to Cambridge. This is not a system in good health.
The future, via Helsinki
A couple of weeks ago I spent a few days in Finland with the FR8 team, interviewing them for Unreasonable. The TL;DR on this clandestine operation is that a group of hugely impressive, wonderfully ambitious twenty-somethings have built a “hacker hotel” in a palace in central Helsinki, providing an unstructured environment for young people (16-28) with the desire to build something the rest of the world would consider insane, surrounded by their peers.
Many of them are university dropouts, some never even applied. There is no curriculum, no classes, no (mandated) equity stake, and no necessary output for the 30-strong cohort in the 12,000m² palace (photo below). This is a maker space at its purest. The team exists to help steer and guide, but not to direct. Building a business out of the programme is fine, but so is building pretty much anything else.
I spent much of the time there wondering if this model was the canary in the coal mine for where higher education goes. Yes – these individuals are outliers. Outcasts, misfits who never fit in in the normal system, and no, this is not new (the Thiel Fellowship, Residency, even YC/EF have been pioneering similar models for years), but the combination of technological advancements and macro conditions suggest this time is different.
The New New
Much like the impact printing had on smashing the hegemony of the church, I believe artificial intelligence will cause a mass fragmentation and unbundling of higher education.
Universities will still exist, yet nowhere near at the same scale – there is just considerably less need for a one-size-fits-all, mass market offering in an era of personalised everything.
In their place will exist new forms of higher education institution – some that are FR8-like, others whose form we can’t yet divine. I fully cede that the FR8 model certainly won’t work for most, and has no plans to as your average teen is nowhere near that agentic, ambitious or internally motivated. Social pressure goes so far, but the firm, guiding hand of the educator goes further.
What I’m interested in is what new forms of experience we’ll cook up in the years to come.1
Growing up
Higher education serves four overlapping purposes:
Knowledge transfer – lectures, reading lists, essays, lab work – the learning bit
Hard skills – you can’t become a (particularly good) surgeon by watching Reels
Credentialing and signalling – the brand on the CV, signalling to the market you’re employable, a proxy for smarts and work ethic
Life-stage levelling – finding your tribe, leaving home, fending for yourself for the first time, part time work – i.e. the wonky, messy growing-up bit
For some the learning is most important (e.g. a doctor or engineer), while for others (generalist arts degree) the learning can actually be secondary to the preparation for adulthood. Credentialing is downstream of the entrance requirements, skills and knowledge learned along the way.
What AI is doing is challenging these four areas in different ways, opening up new potential models as it does. It is posing a major challenge to the previous hegemony of knowledge transfer and, increasingly, hard skills but is yet to (and perhaps never will) touch life stage levelling and network building.
Death to credentials
Yet in posing a major threat to these two skills-based areas, it has in turn shaken the foundations of credentialing systems higher education institutions project to the outside world.
If anyone can produce a competent looking essay or lab report with an LLM, a generalist degree means almost nothing to the world of work. This is especially pointed given we’ve existed for the last ~25 years in a world where around half of UK and US students get an undergraduate degree, and social sciences, business, law and liberal arts make up between 45-50% of all degrees in these nations.
The market has already been pricing this in for years. The FT reported that five years after graduating, business students had an average annual salary of £33,200, less than those with a nursing degree and far behind graduates in medicine and economics (2022-23 data).
With the proliferation of middling universities and degrees, we’ve ended up in a strange world where we use the same credentialing system for a Fields Medal-level mathematician as someone with a business degree from an ex-polytechnic.
What FR8 has woken up to is that new credentialing models are necessary in this changing world – based on the things that hold value in a commoditising working environment. Silicon Valley did half of this work twenty years ago by turning dropping out of university into a higher status activity than staying the course, but this still in part relies on the initial signal of “I got into Stanford”.
Universities won’t die, they’ll be forced back to their core academic purpose – fewer of them, higher quality, fewer degrees, more specialism and more about focusing on those who will actually push academic fields forward.
Alongside this, higher education will unbundle. The FR8 signal is an example of how this could work. Their product, ironically, is more of a competitor to your top level universities, but with an agentic “build it” spin. But given the renewed importance of demonstrating output not process, I can see how this same model could flow downstream into more mainstream offerings.
Signalling will change in line with this. We’ll likely end up with more fragmented credentials, harder to divine to the generalist, but more valuable in their specialist environment. This already exists in vocational fields – most electrical or medical qualifications are utterly unintelligible to the lay-person – but we don’t yet know what this looks like for a more generalist audience where broad categories for white collar work have served some purpose for the majority of this century.
From didactic to agentic
Beyond the signalling is the way these new models could actually work. As we dismantle these tired credentials, the way humans learn must also change. Pure fact repetition, essay writing and repeat lab work served well in a world where basic critical thinking and repetition held the majority of value in most jobs. But as both of these areas become commoditised by models orders of magnitude more capable than the average person (and on an exponential improvement curve), this is no longer fit for purpose.
As the New York Times noted, “a coder is now more like an architect than a construction worker”. This is in part why the “drop out” culture has become so desirable – it’s not just the impatience to start building things, it’s the realisation that the skills taught in a Stanford CS degree no longer match the needs of the real world enough to justify staying the four year course.
So what comes after? Again here, FR8 offers insight.
As intelligence has commoditised, value compounds in being able to do (or orchestrate) work, knowing how to make important decisions at the right time and bringing a human touch to something otherwise cold and digital.
FR8’s model is hyper agentic – empowering and enabling individuals to learn for themselves and drive work forward without external instruction or input. This could be building a propulsion system in space or understanding how proteins fold – the important thing is that the individual is in the driving seat, and is judged on their individual output, not conformity to a mark scheme or exam structure.2
Imagine a specialised wellness centre instead of a sports science degree. Gone is the majority lecture-focus. Instead students learn the necessary scientific underpinnings through hands-on practice, sampling many disciplines (massage, osteopathy, acupuncture etc.) before going deep into a specialism. Human and AI teachers work alongside one another, and peer-to-peer learning and training is built in. Student gain practical experience through industry internships, and emerge from a shorter, sharper course with the academic discipline of before but a much stronger real world grounding.
Peer to peer
Much of these new models exist, just not yet in the mainstream, such as the Ecole42 model for software engineering education. These schools (such as 01Founders in London or //kood in Estonia) pioneer a fully teacher-less model, where students are set up on day one, in person, with a fully digital system where they learn alongside one another and teach other students as they go.
Everything is project based, focused on deliverables, and the model can adapt as the frontier of software engineering changes before it. Most students still need guidance, but the shift from atomised individual learning to active peer collaboration (including AI-assisted peer marking) feels like a logical next step – especially as human labour isn't getting any cheaper.
Fully personalised
One of the fascinating enablers to this transition is the ability of technology to personalise the experience at minimal marginal cost.
It has long been known that personalised learning works. Benjamin Bloom discovered in 1984 that students who received one-on-one tutoring combined with mastery learning performed two standard deviations better than students in conventional classrooms (= the average tutored student outperformed 98% of their peers). Yet tutoring has traditionally been limited to the wealthy due to the high labour costs associated.
The challenge wasn’t in understanding the opportunity, it was in scaling it. There just weren’t enough teachers, hours, or pounds to give every student a personal tutor. And alongside this, cohort sizes inflated as the state became ever more thinly stretched across more universities, dulling the product further.
AI has fully collapsed the teaching cost. One teacher can now do the job of five – more architect than builder. The FR8 cohorts already know it because they’ve done it themselves, but imagine the power of this for the broader economy. By using LLMs as agentic teachers, anyone can learn anything, then act upon it in the real world.
And with this personalisation, much of the associated timelines can collapse as well. Imagine if there were six months or year long hyper-personalised general exploration courses as feeders – with the specialised degree still a possibility as an offshoot, or the opposite – an “action” year following the specialised course that provides the real world experience but with some structure and guidance.
There’s an interesting parallel here in Art Foundation courses but for broader learning. A space for young people to experiment in a less structured environment vs. locking in for a weighty 3-4 year course with limited flexibility.
Friends & feelings
Of the four purposes the university serves, AI poses serious threats to knowledge transfer, hard skills and credentialing. The fourth is what’s left – the human bit, and is arguably the most important part to bring with us as we enter a new era.
The friction around having to make new friends, soft launching adulthood without the real pitfalls of unemployment, the gradual emotional maturing from a fresh-faced fresher into a job-ready graduate, are not to be underestimated, and the worry with speed-running higher education via technology is we lose some of this gradual, developmental phase.
Looking to FR8, I worry less about this. The experience may only be three months, but every cohort member I spoke to spoke of the intense feeling of community, alignment and belonging that was engendered almost immediately upon arrival. I don’t believe this maturation process is as much about length of stay as quality of experience, and so if that can be preserved without the need for a weighty, expensive, long duration degree, I’m all for it.
Let’s talk privilege
Of equal note is the importance of ensuring these new models don’t just serve the well-off or well-connected.
The 17-year-old who already knows what they want to build, has parents or mentors that can vouch for them, and the confidence to apply to a FR8-like programme, will thrive in this world.
The 17-year-old from a working-class town with no professional networks, no role models, no idea what they want to do yet, who needs three years of structured time and a brand-name credential to crowbar their way into the professional class, is the loser of this transition.
Now these technologies are innately democratising – the barrier to entry of using ChatGPT is having a phone or computer – but the credentialing layer is a different question, and there's no guarantee it follows.
It’s worth weighing that while the traditional university was far from perfect, it has long served as a vital engine for social mobility. If the state plays no (or less) role in this future, the new educational model risks replacing university with something that entrenches advantage rather than disrupting it.
And through it all, Catholicism sustains
Oxford University is 930 years old. Harvard has an endowment of $56.9 billion. Like the Catholic Church in 1520, they aren’t going anywhere.
But their hegemony is ending. In their wake a thousand new sects will emerge. Most will be small, many will fail, but a few will become huge new institutions whose forms the medieval mind couldn’t have imagined. This is FR8, 42, Thiel, and the descendants we haven’t seen yet.
This transition will likely be ugly. Mid-tier universities will close, some good ones with them. The models that replace them will be imperfect in new and unfamiliar ways.
But in many ways this future resembles the past more than we might like to consider. Fewer, elite academic institutions, plus a broad long tail of specialised qualifications and centres of learning, whose credentials and signals mean little to the wider world.
The idea of everyone studying a business degree would have meant little to the academic or industrialist in 1900, yet in many ways the world they inhabited was considerably more effective at getting things built, fostering big ideas and learning how to get things done than the one we inhabit today. Food for thought.
I’m Max - I write sporadically about things that interest me, often centred around technology and how it affects our lives, while investing in start-ups via Kindred Capital and Anti Ordinary. Thoughts? Comment below, or just binge my content @ maxbray.xyz
My focus here is higher/vocational education. The school system will change, no doubt. Alpha Schools have shown this already (not without controversy), but pre-16 education is likely to continue looking more like something we recognise for longer, in fact we may well all end up going more Sweden and taking school fully back to pencil and paper as we continue down the path of disintegrating concentration and “social-media-as-smoking”. This is a discussion for another time.
University has always done elements of this output focus – a thesis is a flirtation with individual work in a structured environment – but the output is now largely anachronistic. For the vast majority of students, this essay-based system no longer provides much value.