An Architecture of Ignorance

This article was featured in the Daily Prospect newsletter.

Gas prices were hovering just below $3 a gallon in late February before Donald Trump decided to attack Iran. Due to the closure of the Strait of Hormuz, the national average peaked in early May near $4.50, a 50 percent increase. When Trump capitulated in June and paused bombing, the average price dropped to just above $4. With expectations of protracted conflict ahead, it’s now around $4.40.

We know all of this because gas prices are public and easy to monitor, and because the AAA motor club keeps a well-respected calculator. We also know that fear of measurables—the price of gas, the cost of food, the stock market—is the primary, perhaps the only, discipline on Trump’s anti-democratic instincts. Numbers gave citizens ammunition to apply political pressure. That is what numbers are for. They show us where the problem is, and whether anything we did about it worked.

But the numbers that matter most to a democracy are not as easy to track. Unemployment, educational attainment, what pollutes the air outside a chemical plant; this data needs to be primarily collected and distributed by the federal government. And over the course of his second term, Trump has systematically dismantled this capacity. He is building what we call an “architecture of ignorance,” and its cost is the slow disabling of democratic accountability itself.

One of us is a historian, the other an engineer who spends his days inside the small discipline of observability, the science of inferring a system’s internal state from its outputs. The distinction that discipline draws is between monitoring and seeing. Monitoring tells you the server is powered on. Observability tells you whether it is doing its work as intended. A system that loses observability has not necessarily failed; rather, it has become unknowable. It may be running correctly, or it may have begun to fail an hour ago. From the outside, there is no way to tell. In airplane terms, an altimeter goes dark, and whether the aircraft is at 30,000 feet or 600 is a question that admits no answer until impact.

That is how the systems analyst sees Trump 2.0. Monitors all over the place are going dark.

The historian’s eye runs the other way, toward precedent. Societies that lose the ability to measure themselves rarely register the loss as it happens. They register it afterward, in the accumulated weight of consequences that only become legible once it is too late to do anything about them.

Between the two of us, we are terrified by what we are watching. One of us can see the mechanism failing in real time; the other knows how this kind of story tends to end.

IN TECH TERMS, THE FEDERAL GOVERNMENT operates what an engineer would call a “sovereign stack,” the layered infrastructure of measurement, reporting, and verification that lets a state see itself. The Bureau of Labor Statistics (BLS) measures employment and wages. The Census Bureau counts people. The Centers for Disease Control and Prevention (CDC) tracks disease. Each agency is a sensor, and together they make up the instrument panel through which the state and the electorate read the consequences of policy.

What Trump is doing to these agencies is not deregulation, the marketing-savvy term conjured up by the modern right. It is something narrower and stranger. He is switching off the sensors. The law itself is left untouched, but the conditions the law was written to address can no longer be measured. The statute stays on the books, but the capacity to enforce it quietly disappears.

There is some historical logic to this. By 1992, it had become clear that Reaganomics had not worked. Ronald Reagan’s three-part economic strategy of lowering taxes (especially on the wealthy), slashing social programs, and increasing defense spending turned out to be little more than class warfare, wealth transferred from the poor to the rich. It is a well-worn parlor trick among economic historians to pull up graphs of quality-of-life indicators across Western democracies—life expectancy, the size of the middle class, educational attainment—put Reagan’s face on the year 1980, and watch the United States diverge from its peers on measure after measure: income inequality, child mortality, access to health care.

The Prospect in your inbox

Analysis that goes beyond the noise with in-depth, progressive reporting.

No paywall, no programmatic ads — just the reporting. Unsubscribe anytime. Submitting your email signs you up for The Daily Prospect, our weekday morning newsletter.

When the numbers tell that story, the numbers become the enemy, and the political problem becomes how to hide them. This is not a motive we have to infer. Project 2025: Mandate for Leadership, the governing blueprint of Trump’s second term, named the statistical agencies as targets for this very reason.

Project 2025, for instance, called for the Census Bureau and its sister agencies to be brought under direct political supervision to ensure “alignment” with administration principles (p. 664). The vocabulary is reminiscent of corporate restructuring: “alignment,” “efficiency,” “reducing redundancy.” But the meaning is political. An apparatus built to measure the government’s performance is being handed to the people whose performance it measures.

Three examples show how it works.

Take the National Assessment of Educational Progress, often called the Nation’s Report Card. Its value lies almost entirely in a trend line spanning half a century. It is what lets us see the long-tail effects of the No Child Left Behind education law, the disruption of COVID-19, and the arrival of digital instruction. In 2025, the administration gutted the National Center for Education Statistics, the agency that administers it. Because of that, the 2026 assessment is dead, and the 2028 assessment is likely dead, too. Even if Democrats take office in 2029, they cannot go back and measure what fourth graders knew in 2026. That cohort will have aged out. We will simply never know what happened to a generation of students during these years.

Or take the Environmental Protection Agency, which in September 2025 suspended its fence-line monitoring requirements for coke plants (where fuel is produced by heating coal or petroleum) and chemical manufacturers. The suspension did not let those facilities pollute; they were already permitted to pollute within limits. What it suspended was the verification that the limits are being observed.

So when leukemia clusters surface a decade from now in the ZIP codes around St. James Parish, Louisiana, or Brazoria County, Texas, there will be no baseline record of benzene and 1,3-butadiene concentrations to establish causation. The families will know their children are sick, and they will suspect the plant, and they will probably be right. They just won’t be able to prove it in court, where causation gets established through records that, in this case, nobody was required to keep.

Or take labor statistics. On August 1, 2025, the BLS commissioner was fired within hours of a jobs report the administration found unfavorable. The stated justification was a dispute over methodology. The real explanation was unearthed three months earlier. On May 16, Moody’s downgraded U.S. sovereign debt, citing the fiscal trajectory of the administration’s tax cuts. The 30-year Treasury yield crossed 5 percent, and the cost of servicing the national debt spiked. The administration had received a signal it could not answer through fiscal policy. That signal had a price, paid in the bond market in basis points. So what did the administration do? It went after the signal.

This is how credit, both fiscal and reputational, gets destroyed, not through default but by making the underlying data impossible to trust. The fiscal damage alone is bad enough. The global financial system runs on the assumption that the United States is a predictable counterparty whose self-reported numbers are reliable. Compromise that assumption and the risk on U.S. assets goes up. Borrowing costs rise and the dollar weakens, while pension funds and sovereign-wealth allocators quietly reprice how much of us they want to hold.

Trump seems to believe this is reversible, but the bond market does not work that way. Credibility, once compromised, takes years to earn back.

The courts are supposed to be the backstop to the destruction of the sensors, and here the design was most precise. The Administrative Procedure Act of 1946 requires agencies to justify their decisions with evidence. When an agency makes or repeals a rule, it must show that the decision was not arbitrary, pointing to data, public comments, or studies to make its case. For most of the postwar period, this was a genuine constraint on executive power, and most presidents accepted it. One historian called the APA “the quasi-constitution of the modern administrative state.” The EPA could not ban a pesticide without showing the science, or unban one without showing the science had changed. A court always had something in front of it to review.

But what happens when the agency stops collecting the data in the first place? A court reviewing an empty record cannot find that an agency ignored data that was never there. The presumption of regularity, the doctrine that officials are presumed to be acting in good faith, stops being a courtesy and becomes a shield, because there are no facts left to pierce it.

The older forms of administrative defiance all left a record to fight over: evidence ignored, suppressed, disputed. The architecture of ignorance leaves nothing to fight over, because the evidence never enters the world at all.

THERE WAS, UNTIL RECENTLY, one more check in place: the independence of certain federal agencies themselves, whose chief administrators were shielded from presidential removal except for cause. But on June 29, in Trump v. Slaughter, the U.S. Supreme Court handed the president expanded power to remove that leadership from agencies long held to be independent. With the Court already inclined to enlarge executive power at nearly every turn, Slaughter knocked out one of the last structural reasons an official could give for collecting a number the president would rather not see.

And here is where it gets frightening. By making sure the numbers are never collected in the first place, Trump is producing something far more durable than any policy win. He is salting the earth for whoever governs next. We call this running cost “epistemic debt”: the obligation incurred every month a measurement that should be taken is not. Financial debt can at least be paid down later. Epistemic debt cannot. The data not collected in June 2026 cannot be collected in June 2030. Each disbanded research team is institutional knowledge gone, and each canceled study is a thread cut out of the historical record.

The historian here reaches for precedent and finds a spectrum. At the benign end, Major League Baseball brackets its “steroid era” with an asterisk. The home run and RBI records still stand, but they carry a permanent notation signifying that the numbers cannot quite be trusted because the players were juiced with performance-enhancing drugs.

Further along the spectrum, violent regimes have destroyed records already taken. For instance, when the 1937 Soviet census counted millions fewer people than Stalin’s government had claimed, the results were buried and the census-takers arrested, many of them shot. Along similar lines, other authoritarian regimes have simply stopped producing numbers. As its economy stumbled, China removed hundreds of economic indicators from public view, among them land sales, foreign investment, and youth unemployment.

What is being built here in the United States sits at this far end of the spectrum.

In politics, what replaces measurement is performance. Maybe it always did, which might be why in 2024 Joe Biden was condemned for high inflation in a year when inflation was in fact relatively low.

The Prospect in your inbox

Analysis that goes beyond the noise with in-depth, progressive reporting.

No paywall, no programmatic ads — just the reporting. Unsubscribe anytime. Submitting your email signs you up for The Daily Prospect, our weekday morning newsletter.

But now things are considerably worse. Without reliable data, governance can no longer be judged by results, and all that is left is how a leader makes people feel. This is governance by vibes, what we call “affective sovereignty,” where legitimacy comes not from outcomes but from the projection of strength. A president can call the economy the greatest in history, and without a credible BLS the claim floats free of reality. The effect, whether or not anyone designed it that way, is to move politics entirely into the realm of the aesthetic, which is precisely where a kleptocracy is safest.

Rebecca Kelly Slaughter, the FTC commissioner the Slaughter case was named for, put it best: “Today’s ruling makes it possible for presidents to fire watchdogs who won’t put politics over principle, and replace them with lap dogs.” The most frightening part is that none of it requires a conspiracy, or even much coordination. It requires only that the people who benefit from the blackout notice the architecture is available, and use it.

It should be said that some agencies continue to supply useful data. The BLS jobs report released last Friday, the final measurement before the election, showed hiring grinding to a standstill and unemployment rising slightly. It was not what Trump would have wanted out there. But this kind of inescapable, unspinnable data is increasingly becoming exception rather than rule.

Recovery, if it comes, will be uneven, and it is worth being honest about what can be salvaged and what cannot. Where staff and infrastructure survive, a future administration can start measuring again. And fortunately a trend line with a hole in it still tells you something. But where the infrastructure itself has been dismantled, there is no quick restart. Institutional knowledge takes decades to rebuild, and people will be hurt in the meantime. And past a certain number of gaps, there is the inevitable suspicion that any given number is just an artifact of whoever held power when it was taken. A sensor that only reports good news is worse than no sensor at all, because you can no longer tell a change in the world from a change in the instrument.

No matter what happens in 2028, an administration committed to evidence-based governance will inherit a shattered dashboard. There will be no baseline for the toxins in the air, and no honest measure of what a generation of students lost. The state will have gained the power to assert and to impose, with no feedback loop left to tell any of us how things are actually going. The rest of us will live with the consequences. The record by which those consequences might have been known will never exist.

The post appeared first on The American Prospect.