Zombie Institutions

In 2017, the FCC asked the public what it thought about net neutrality and received more than 22 million comments. New York's attorney general later found that about 18 million of them were fake. Nobody needed a language model to fake them.

Public comment rests on one assumption: one comment, one person. That assumption died years ago, and the process still runs. It collects the comments, counts them, and reports the totals as if they meant something. That's a zombie institution: a structure that keeps enforcing itself after its reason is gone.

In an earlier essay, I argued that two observations explain why movements against power so often end up serving it:

Structures outlive their reasons.
Empty structures fill with whatever energy is nearby.

Capture comes fastest to institutions with no finish line, and it's nearly invisible from inside. This essay asks what that means now, while a new information technology outruns every institution we have.

Technology first, institutions later

Every new information technology has outrun the institutions meant to govern it. Print arrived around 1455. Europe spent much of the next two centuries in religious war, and the Royal Society, an institution for deciding which printed claims to trust, arrived about 205 years after the press. Electricity moved faster. Underwriters Laboratories, founded by fire insurers, came twelve years after the first power station. Finance got the SEC within five years of the 1929 crash.

The order is always the same: technology first, institutions later, and a crisis sets the timing. When the crisis comes, it passes whatever is ready. Most of the institutions that crises create already existed as proposals.

The medium writes back

Writing moved memory outside us. Print moved copying. The internet moved distribution. Models move the reading and the writing.

We haven't absorbed the last one. The internet is more than fifty years old, and we still don't agree on what it did to democracy, let alone what to build in response. The fake comments of 2017 were cheap to make and easy to catch in bulk. The next ones will be fluent, specific, and hard to tell from people. Every institution that assumes a human on the other end, reading or writing, is about to find out whether that assumption was load-bearing.

The longer clocks

Some people think institutions have lifespans too.

John Glubb argued that empires last about 250 years, roughly ten generations. The United States turned 250 this year.
William Strauss and Neil Howe described a crisis era, about once a lifetime, that clears out old institutions and builds new ones. The last one ran from the 1929 crash through the Second World War. Their schedule puts the next one in our time.
Peter Turchin, working from much better data, forecast in 2010 that American political instability would peak around 2020.

None of these is a law. Glubb's sample was hand-picked, and the Fourth Turning is closer to a mood than a model. Turchin has the strongest data and makes the most modest claims. Treat all three as questions. But if they're even partly right, we're adding the most powerful information technology in history to institutions that were already near the end of a cycle.

Three questions for now

What are our zombie institutions? Which institutions still enforce rules whose reasons are gone? A few candidates, offered as questions, not verdicts:

Public comment assumes one comment means one person.
Credentials assume that writing an essay proves you can think.
Copyright was written in 1710 for a world where copying was expensive.
Election law tracks money spent, while influence moves through feeds no one audits.

Some of these are Chesterton's fences, with reasons we've forgotten. Some are zombies. From inside, they look the same. Which are which?

What are our institutions for? Capture comes for institutions with no finish line. Most of ours can't say what winning would look like, and neither can most of the companies building AI. "Benefit all of humanity" has no finish line. An institution that can't state its goal optimizes whatever it can measure, and whoever writes the score decides what gets optimized.

Meanwhile, the largest flow of energy in the economy runs through four companies, which will spend about $720 to $745 billion on AI this year. Institutions with unclear goals turn slowly and involuntarily toward the largest source of energy nearby. Where is that turn taking ours?

How does power work in the new topology? Our institutions were built for power that came in hierarchies, stayed inside borders, and depended on people. Power now runs through networks: whoever controls the feed, the compute, and the models.

Antitrust law measures prices, and the products are free.
Election law tracks spending, and influence moves through recommendation.
Laws stop at borders, and networks don't.

The deepest change is the last check of all. Power has always needed people: soldiers, workers, taxpayers. Václav Havel's greengrocer, putting the regime's slogan in his shop window, had power because the regime needed him to put it there. What happens to the power of the powerless when the powerful no longer need anyone's participation?

What to do now

For one person:

Run the capture test on everything you belong to, including yourself. Would you be relieved or threatened if your stated problem were solved tomorrow?
Find a rule you enforce but don't believe, and check whether it still has a reason.
Choose what you build, buy, back, and refuse. That's where most people's agency actually lives.
Keep a practice of return. Your own clarity will drift too. Plan to come back to it.

For a group:

Build institutions before the crisis. Crises pass whatever is ready. Have something ready.
Write capture defenses in at the founding. Alcoholics Anonymous refuses outside money, stays nonprofessional, and owns nothing, and it still runs on those rules nearly a century later. Give every new institution a finish line, limits on outside money, rotation, and a plan to re-found.
Build sense-making that bridges. Systems like Community Notes surface what people who usually disagree both accept, instead of what spreads fastest.
Keep capability plural. Totalitarian rule needs a monopoly on information and force. The counterweight is plural capability and shared ways to check what's true.

The questions we have to answer

Which of our institutions are zombies, and which are fences whose reasons we've forgotten?
What is each institution we depend on for, and what would finished look like?
Where is the energy flowing now, and what is it pulling everything toward?
What checks power once power no longer needs people?
Is any of this something we can change, or only something we can slow?

None of this is a forecast. The technology doesn't pick the ending. People do, in what they build, buy, back, and refuse. Underneath every question here is the one each crisis asks: when it comes, what will be ready?