The Four Phases of Institutional Collapse in the Age of AI

How democracy, expertise, and institutional memory are eroding during technological change - or @grok is this true?

🎂It is the one year birthday of In This Economy? Thank you so much for all the support and kindness. If you enjoyed the book, please leave a review, share it with a friend, all that good stuff. Paperback available now!

Big Tech, Broken Institutions, and the Trouble Ahead

A federal court has deemed Donald Trump’s tariffs illegal, stating essentially that

1. The President isn’t a king

2. The government now has to pay back tariffs.

Trump can still do certain types of tariffs, like the ones under Section 232 and 301 which require more process (the tariffs on steel, aluminum, autos, pharmaceuticals, electronics, and lumber).

The Justice Department immediately appealed. It looks like it will go in front of the Supreme Court, which will be a big test of the institutions of the 21st century. But legally, Trump got TACO’ed - there is now a ruling that indeed, Trump Does Always Chicken Out. What will he do next? Stephen Miller already called it a ‘judicial coup’.

And that’s what I want to talk about today - what happens when we experience massive technological change at the same time that the institutions meant to help manage that change are being dismantled or threatened or ignored?

The Invisible Rules

The other day, I was biking up a steep (Steep!) grade, and a girl - holding two phones - stepped into the middle of the road and just… stopped. Right in front of me! Not a flinch. I swerved and kind of laughed and then thought: Wait, is this real? Or am I in some kind of dream?

Because it kind of feels like that lately. Like the invisible rules that used to hold everything together like the rules about paying attention, about looking out for one another, about knowing where you’re supposed to be are just… gone.¹. No one is where they’re supposed to be (at least, according to the Societal Rulebook™️).

And that is sort of how everything feels right now? This constant weaving and waning through traffic littered with people not really paying attention or sucked in their alternate phone reality in ways that makes you want to shake your fist to the sky.

Meanwhile, I keep seeing these little delivery robots, zooming down sidewalks, sitting patiently at crosswalks, delivering a single KitKat or something to someone in the middle of Mar Vista. They are intentionally cheerful (the googly eyes) scouts from a future that’s already here, waiting for someone, anyone to put in an order.

Meanwhile, humans sit in nearby parking lots, scrolling phones in idling cars, also waiting for someone, anyone… The robots know exactly where they're going. The humans… don't?

Robots don’t need sleep or a 401k. Pictured here outside a weed store.

The First Civilization to Demolish During Transition

We are living through a major economic transition. I think we all know that. Every major technological shift throughout history demands institutional adaptation - labor laws, public education systems, financial regulations, safety standards, etc.

But today feels… ~different².

We don’t really know what we’re transitioning into. There’s no dramatic unemployment crisis yet³ (in fact, Klarna and Duolingo are walking back some of their AI-ification). But there’s a broad, creeping unease like everyone senses something big is happening, and no one feels ready… This might be the first major technological shift where we aren’t actively building new institutions to manage the change, but instead we’re dismantling the ones we already had.

• The government is (was?) actively being dismantled, with Elon Musk now stepping down from DOGE after doing many harmful cuts.

• Robert F Kennedy, Health Secretary, is citing fake sources in his Make America Healthy Again report according to NOTUS as he attempts to turn modern medicine back into leeches and letting or whatever

• Sean Duffy, Transportation Secretary, went on Fox and said “if you're liberal, they want you to take public transportation ... the problem is that it's dirty.” It is his job to improve public transportation.

Meanwhile:

• American funding for hard sciences has fallen by two-thirds this year.

• The Trump administration is pausing student visa interviews in order to vet students’ social media (not very free speech, and certainly not stapling a green card to every college diploma of them)

• Small rust belt college towns are collapsing.

Exactly when we need maximum institutional competence to manage algorithmic coordination of the economy, we're destroying the institutions that could provide that. Like, this guy is in charge of the technology that he says is going to destroy us! We need some sort of guardrail!

We have the true near death of expertise at the same time that AI is apparently going to take all of our jobs. There are plenty of articles tracking the visible changes going on - the self-enrichment, the impatience, the attack on knowledge - but very few are reflecting on the technological changes that are happening in parallel.⁴

Institutional Competence During an Algorithmic Transition

I think it’s happening like this -

This isn’t a linear story arc or a roadmap of change etc, just a rough progression. These phases overlap, feed into each other, contradict themselves. But they help explain what’s going on right now. Why everything feels like it’s… eroding.

Phase One: Trust Erosion

Institutions remain, but people stop believing they work for them.

We are here! More and more, human labor is coordinated by algorithm: CRMs, KPIs, valuation models, algorithmic feeds. Every software-based industry is part of this number-centric system (which is many industries). Not inherently bad! But it leaves a lot of people with the job title Biological API, or humans executing digital instructions.

At the same time, institutions are being politically captured (apparently universities have to align with the administration). Science funding gets cut for ideological reasons. Economic policy becomes kayfabe and theater. Etc. When everything becomes political, expertise becomes just another opinion to be negotiated rather than a constraint to be respected.

What breaks here is trust (the thing I never stop writing about). Institutions remain, but people stop believing they work for them.

Phase Two: Knowledge Erosion

Institutions stop functioning as engines of competence transfer.

We're already seeing the systematic dismissal of technical knowledge - Tom Nichols wrote a whole book on it, so none of this is new. Climate science becomes "liberal bias," and epidemiology becomes "a control agenda" or whatever.

At the same time, unemployment among college graduates is rising faster than for other groups (at the same time that student loan payments are starting back up). There are very real economic consequences here. The value of higher education is being questioned while the cost remains the same (or rises).

And then there’s the labor market. LinkedIn's Chief Economic Opportunity Officer Aneesh Raman recently compared the surge of automation to the collapse of US manufacturing in the 1980s in a NYT Op-ed.

We saw what happened in the 1980s when our manufacturing sector steeply declined. Now it is our office workers who are staring down the same kind of technological and economic disruption.

Breaking first is the bottom rung of the career ladder.

The change is slow. An IBM survey shows that 3 in 4 AI initiatives fail to deliver promised ROI (perhaps because implementation costs exceed immediate benefits). But entry-level positions where young people once cut their teeth (like debugging simple code, basic legal research, administrative tasks, or what I did in my first job, putting numbers in a spreadsheet) are increasingly (expected to be) handled by AI.

That’s the rung Raman was talking about. And if that rung disappears, what is left for the people trying to climb?

This shift creates a two-track economy: fully automated work (data analysis, customer service, financial calculations) done by AI versus remaining human domains (complex negotiation, creative problem-solving, emotional support) that can be done by humans.

But to fully enter Phase Two here and takeover remaining human domains, the questions have to get weirder and weirder and then you feel like you’re trapped in some seventh circle of San Francisco hell - with questions like “How do you scale empathy? How do you metrics-ize the kind of relationship-building that really drives sales? How can we operationalize the human spirit?”

So: the technical knowledge is being dismissed, the early career pathways are collapsing, and we’re now trying to automate what’s left of the human stuff too.

What breaks here is knowledge transfer. Institutions stop functioning as engines of intergenerational competence and we lose people entering the system who might have fixed it.

Phase Three: Capacity Erosion

Rebuilding when no one can

Here, institutions break down physically and economically. This is decay beyond job loss. Some of it is already happening. The Wall Street Journal has an incredible piece on the bust of small college towns in the face of the demographic crisis that I mentioned earlier - focusing on Macomb, Illinois, where Western Illinois University enrollment has dropped 47% since 2010.

These are stories about fertility rates⁵ and federal funding cuts, and what happens when (1) the population ages and (2) the university model gets challenged. Automation has no answer for this. As the WSJ reports -

Of students choosing to enter college, more of them are aiming for prestigious universities, believing those diplomas will get them better jobs, said Richard Vedder, an economist at Ohio University who studies higher education. “It’s a flight to quality,” he said.

But, this won’t work, because the US government is actively trying to destroy universities too, revoking student visas for Chinese students and suspending federal funding, among other things. China produces almost half of the world’s top AI researchers and nearly 40% of AI talent in the US is Chinese. Many who get PhDs choose to stay in the US after graduating.

We benefit from these systems, and destroying them is strategically incoherent.

What breaks here is capacity. Even if we wanted to build new institutions, there's too much political infighting and too little focus on functional outcomes. You can’t rebuild if you’ve dismantled the builders.

Phase Four: Algorithmic Substitution

The machine doesn’t just do the job. It becomes the institution.

Once trust, knowledge, and capacity erode, algorithms replace. This is about algorithmic systems bypassing democratic institutions entirely.

This could happen. It would be algorithmic systems bypassing democratic institutions entirely. Decision-making, distribution, prioritization all offloaded to markets, engagement metrics, and predictive software.

This is why Trump's crypto dinner with the weird rubbery steak matters more than just dollar amounts - it represents the emergence of alternative power structures (a version of the parallel economy that I wrote about a few months ago). Access is granted through memecoins and vibes! Not competence.

Markets become more important than administration. Automation displaces expertise. Young people prefer influencers over institutions. This is the rise of structures that don’t require institutional competence at all, but instead require vitality and optimization and crowd psychology.

And for better or worse, Trump is the first human-algorithm hybrid president - governing via Truth Social truths, bond market reactions, and direct market signals. A feedback loop in a suit.

What breaks here is a version of democracy. Algorithms bypass institutions, but institutions were our only mechanism for managing complexity at scale.

Political Optimization

Part of our problem right now, as many have detailed, is that our institutions are designed for the 20th-century (stable demographics, predictable technology, contained information flows) and we really don’t have that structure anymore - we have 21st century realities, like exponential change and global information networks.

Harold Robertson has an interesting piece from 2023 in Palladium⁶ titled Complex Systems Won’t Survive the Competence Crisis that touches on the increasing complexity of everything and the decreasing competency of people involved:

In straightforward work, declining competency means that things happen more slowly, and products are lower quality or more expensive.

In complex systems, declining competency results in catastrophic failures.

Complex systems are complex! And when they get stupid, three things happen: redundancy consumption (drawing down institutional reserves), complexity reduction (making things simpler but less capable), and risk externalization (pushing problems to other systems). But eventually, things stop working. And the competence crisis gives AI a market demand, a reason to exist, an economic justification to replace existing institutions.

Implementation costs are still high, human oversight is still needed, most people still prefer talking to a human, there are regulatory barriers in place, etc. The timeline is longer than the hype suggests, but it’s still there.

We have a true, bright test of whether complex technological civilization is compatible with democratic political optimization.

But we’re applying 20th-century frameworks like job retraining, education reform, social safety nets to a fundamentally different kind of economic transformation. These assume the basic structure of human-necessary work will persist, just in different forms. It might… not? I think it will. But we don’t seem to be prepared for the alternative (as many have discussed).

And we may be approaching a phase transition where the underlying structure changes state entirely. The question now is how to navigate the transition to whatever comes next, which is where people like Tyler Cowen are really helpful. Previous technological transitions succeeded because societies built institutional capacity to manage change. The industrial revolution created labor protections, public education, and financial regulations.

We need equivalent institution-building for the algorithmic age, but first, we need to stop demolishing the knowledge infrastructure we already have! This means doing obvious things like protecting science funding, maintaining international academic relationships, and preserving institutional memory even as we adapt to new realities.

I often think of Jerome Powell and what he is thinking right now and how hard his job is at the moment. His advice is have integrity and guard it carefully. As he said in his Princeton commencement address “Fifty years from now, you will want to be able to look in the mirror and know that you did what you thought was right, in every part of your life” He is right. The machines know where they’re going. The humans have been waiting in the parking lot for years. Maybe it’s time to decide where we’re going too.

Thanks for reading.

1. Old man yelling at cloud meme

2. Cue the ‘this time it’s different bell’

3. Young people are experiencing the effects of a slowing economy and technological change, which I’ll get to later in the piece

4. Tyler Cowen has an incredible piece with the Free Press titled ‘AI Will Change What It Is To Be Human? Are We Ready?’ and I don’t think we are. I don’t know if we know how to be.

5. As the WSJ reports “In 2007, the number of U.S. births peaked at 4.3 million and has been falling almost every year since”

6. Harold diagnoses the competency problem as diversity initiatives which I don’t agree with at all.