The Hidden Cost of Running a Society on Stale Data

What if the real bottleneck is not intelligence, but latency?

We keep telling ourselves that the big economic and institutional fights are about ideology: left versus right, state versus market, independence versus control. But there is a more technical and more consequential question hiding underneath all of that: who gets to make decisions when the world is changing faster than the institutions that govern it?

That question shows up everywhere at once. In labor markets, only a fraction of public sector work appears to be fully automatable, but the exposure is concentrated in the boring, administrative layers: meetings, bookings, forms, records. In monetary policy, a small committee still sets the terms of the global cost of capital using month old or worse data. In industrial policy, governments are rediscovering that if they are going to take risk with public balance sheets, they may as well ask for upside. And in biology, models that were originally built for language are now generating proteins that can meaningfully change how cells age.

The common thread is not AI, not rates, not industrial policy, and not biotech. It is latency. The world is shifting from systems that depend on slow human interpretation to systems that can ingest reality in near real time, produce predictions quickly, and sometimes act faster than human committees can meet.

That shift is more than a productivity story. It changes the logic of institutions themselves.

---

The old model: humans as bottlenecks for reality

Most large institutions were built around a simple assumption: reality is too noisy, too expensive, and too hard to measure continuously, so a small number of trusted people must sample it periodically and decide what to do.

That model made sense in an era of paper trails, quarterly reports, and slow communication. It makes less sense in a world where payroll systems can emit live labor data, blockchain ledgers can publish macro indicators, markets can reprice assets in milliseconds, and machine learning models can infer structure from patterns no person would ever detect by hand.

This is why the public sector labor data is so revealing. If an average public worker works about 8.3 hours a day and roughly 3.4 hours are exposed to generative AI, the obvious conclusion is not that government jobs disappear overnight. It is that a huge amount of institutional time is spent on coordination, documentation, and translation. The tasks most exposed are not the glamorous parts of public service. They are the administrative pipes: scheduling, form processing, record keeping, booking, meeting administration.

That matters because these are not peripheral tasks. They are the glue that allows institutions to move at all. If you can automate the glue, you do not just save hours. You change the speed at which the whole institution can respond.

The deepest disruption from AI is not replacing experts. It is shrinking the time between signal and action.

Once you see this, a lot of other debates snap into focus. The Fed is not merely a political battleground. It is an institution built to interpret a world that now sends better signals faster than the institution can digest them. Commercial real estate is not just a financing problem. It is a balance sheet correction delayed by years of cheap capital. Corporate bankruptcies are not just bad news. They are the return of decision making to markets after a long period of artificial delay.

The issue is not whether institutions should exist. The issue is whether they still deserve to be the slowest thing in the system.

---

When delay looks like stability, and stability becomes distortion

There is a seductive idea in economics and governance that delay is wisdom. Wait. Observe. Insulate. Protect against panic. This instinct is not irrational. Independence, long terms, and procedural friction can prevent obvious abuse.

But delay has a hidden price: it can convert adaptation into distortion.

Take monetary policy. A handful of governors meets monthly, examining data that may already be stale, revised, or incomplete. In a $130 trillion global economy moving at digital speed, that is a remarkable setup. It is like trying to steer a race car by looking in the rearview mirror while other systems around you are already using live telemetry.

There is a legitimate reason to want distance from politics. If short term pressure controls rates, every election becomes a bid to manipulate the cost of money. That leads to the classic trap: lower short term rates, more borrowing, more demand, more inflation, and eventually higher long term yields and a heavier debt burden. Independence is supposed to protect the long game.

But independence only works if the institution is actually processing reality better than the alternatives. If the signal is poor and the lag is large, the institution becomes less like a stabilizer and more like a source of accumulated error.

The 2021 inflation episode is a perfect illustration. A transitory narrative bought time. Time, in that case, was not neutral. It allowed easy money to persist long enough to feed a bubble in startups, real estate, and speculative capital formation. Then came the violent correction, because the system had delayed adjustment until the necessary response had to be much sharper.

That is the paradox of stale decision making: the longer you wait to recognize reality, the more abrupt reality becomes when it arrives.

Commercial real estate is doing the same thing at asset level. For years, sponsors and banks extended and pretended because neither side wanted to recognize losses. Refinancing now means higher rates, lower valuations, and painful equity gaps. The building was not magically healthy during the period of delay. It was simply being carried by cheap financing and accounting patience. The eventual foreclosures are not a new crisis. They are the bill coming due.

Corporate bankruptcies tell the same story. A surge in failures may look ominous, but in a deeper sense it can be a form of correction. Years of zero rates and abundant capital kept weak businesses alive. Cheap money acted like oxygen for companies that should have exited earlier. When the reservoir runs out, the economy begins to sort again.

In a distorted economy, bankruptcy is not always a symptom of decline. Sometimes it is the first honest signal in years.

That is why treating every increase in failure as evidence of dysfunction is too shallow. In many cases, failure is the system recovering its ability to distinguish strength from survival.

---

A new rule for state power: if you take risk, take the upside too

Once you accept that the world is increasingly governed by real time signal processing, a second question follows naturally: what should governments do when they intervene in markets?

The old answer was simple: give grants, issue loans, provide guarantees, and hope the public gets the benefit indirectly through jobs or strategic resilience. But if the state is already taking balance sheet risk on behalf of the public, why should it not also capture some of the upside?

This is where industrial policy becomes less ideological and more architectural. If a government uses its balance sheet to support strategically important firms, especially in sectors like semiconductors, pharmaceuticals, or critical materials, then pure subsidy is a crude instrument. It socializes downside and privatizes upside. That is not capitalism, and it is not quite public stewardship either. It is one way traffic.

Equity changes the game.

A public stake, even a passive one, reframes the relationship. Instead of saying, “We will give you capital because your success matters to us,” the state says, “Your success matters, but so does the public’s participation in that success.” That is a stronger moral argument and a better fiscal one. It creates transparency, preserves market discipline, and gives citizens a claim on the value created by strategic support.

But equity by itself is not enough. The real design problem is governance. If the state collects shares, where do those shares live? On the federal balance sheet with no strategy? In a new sovereign wealth fund? Or inside an existing public trust with a clear mandate?

This is not just an accounting question. It is a theory of public ownership. If gains are scattered, politicians will spend them. If gains are pooled without purpose, they will be mismanaged. If gains are ringfenced with explicit rules, they can serve long term public objectives.

The most interesting idea here is to treat public equity not as a new spending pool, but as a capital base for future taxpayers. That is a much more disciplined vision. It says the state should not use strategic stakes to feed present day political appetites. It should use them to build an asset base that compounds for the people who will carry the burden later.

In other words: if government is going to behave like an investor, it should also behave like one in the hardest sense, by protecting the principal and letting compounding do its work.

---

The same revolution is happening in biology: better signals, faster iteration

The most surprising connection in all of this may be biotechnology. Models trained on language and sequence data are beginning to generate new proteins that materially improve cellular rejuvenation processes.

Why does that matter in an essay about institutions and economic policy?

Because biology is now learning the same lesson as macroeconomics: if you can represent a complex system as data, you can search it faster than intuition ever could.

A cell does not age because of one dramatic failure. It ages because gene expression networks drift, repair processes weaken, and accumulated noise compounds. The idea behind rejuvenation is not to make the cell into something alien. It is to restore its operating state. That is remarkably similar to what good institutional reform tries to do. You do not replace the whole system. You reset the parts that have drifted too far from function.

What is striking is that an LLM can help design better variants of the proteins involved in this process. It can search enormous sequence spaces, propose candidates, and accelerate experiments that would otherwise be too combinatorially difficult for humans alone. The deeper lesson is not that “AI is powerful.” It is that once a domain becomes legible as patterns in data, the speed of discovery increases dramatically.

That is true in protein design. It is true in lending. It is true in labor administration. It is true in monetary signaling. The information bottleneck moves from human interpretation to machine orchestration.

The analogy is useful because it reveals something institutions often miss: many systems are not failing because they are badly designed in the abstract. They are failing because the tempo of the system no longer matches the tempo of the environment.

A committee that meets monthly cannot fully govern a market that reprices continuously. A bureaucracy that processes forms manually cannot serve a public that expects instant service. A subsidy program that takes years to realize gains cannot compete with capital markets that react in days. A drug discovery pipeline that relies only on human intuition cannot keep up with models that can explore billions of variants.

The cure is not to remove humans. It is to place humans where judgment matters most and let machines compress the rest.

---

The real institutional test: can you convert friction into feedback?

This gives us a better framework than the usual arguments about deregulation versus control.

Think of every institution as a feedback system with three layers:

1. Signal acquisition: how quickly and accurately the institution sees reality.
2. Interpretation: how it turns that signal into a decision.
3. Transmission: how fast that decision changes outcomes.

Most institutions are strongest in the second layer, at least in theory. They think they are wise because they deliberate. But if their signal acquisition is slow and their transmission is hampered by procedural drag, then the wisdom is mostly theatrical.

That is why AI changes governance more than it changes work in the narrow sense. It compresses layer one and layer three. It lets us see more and act faster. But it also raises a profound question: if the machine can process the signal faster than the institution, what is the institution for?

The answer cannot simply be “to preserve itself.” It has to become a place for judgment, legitimacy, and constraint. Humans should define the objectives, the rules, and the moral boundaries. Machines should help execute within those boundaries.

This is the hidden connection between public sector automation, Fed criticism, public equity stakes, and protein design. They are all examples of a world where the cost of lag is getting unbearable. When lag is high, power accrues to whoever can delay the longest. When lag falls, power shifts to whoever can update fastest.

That is why the most valuable public and private systems of the next decade will not be the ones with the biggest staffs or the most elaborate procedures. They will be the ones that can turn friction into feedback without losing legitimacy.

---

Key Takeaways

• Look for latency, not just inefficiency. When a system feels broken, ask how long it takes to notice reality, not just how badly it behaves.
• Separate judgment from routine. Humans should focus on goals, tradeoffs, and accountability. Machines should handle repetitive coordination and pattern extraction.
• Treat public risk like public capital. If the state takes downside risk to support strategic sectors, it should design a transparent path to share in the upside.
• Assume delay creates bigger corrections. Whether in rates, real estate, or bankruptcies, postponing adjustment often makes the eventual correction harsher.
• Update institutions to the speed of their environment. A monthly committee cannot govern a real time market using stale data and expect precision.

---

The future belongs to systems that can update without panicking

The temptation in every era of disruption is to overcorrect. Some people will say automation means abolish institutions. Others will say preserve institutions exactly as they are. Both instincts are wrong.

What we actually need is a new standard: institutions should be judged by how well they update. Not how slowly they resist change, and not how recklessly they chase novelty, but how well they absorb new information, make a decision, and preserve trust while doing it.

That standard applies to monetary policy, labor administration, industrial strategy, and even medicine. In each case, the winner is not the actor with the most authority. It is the actor with the best feedback loop.

So the deepest shift underway is not that machines are becoming smarter. It is that the cost of being wrong for too long is collapsing.

And once that happens, every institution faces the same choice: become faster, become more precise, or become obsolete.