The Quiet Refounding

On the morning of January 20, 2025, while the world watched an inauguration, an executive order was signed establishing the Department of Government Efficiency. It was not on the front page. The cameras were elsewhere — on the ceremony, on the crowd, on the transfer of power that every civics textbook teaches us to watch. DOGE was announced as a cost-cutting initiative: a Silicon Valley efficiency play applied to the federal bureaucracy. The language was deliberate. Efficiency. Modernization. Streamlining. Who could oppose efficiency?

Within eighteen months, AI tools with names like “SweetREX” were deployed at the Department of Housing and Urban Development, scanning federal regulations and recommending which ones to delete. AI systems were monitoring federal employees’ communications for signs of disloyalty. Executive orders had preempted state-level AI regulation, centralizing control of the most consequential technology in human history. Another executive order had integrated AI into national security decision-making systems. And a separate order on digital financial technology had simultaneously banned central bank digital currencies while promoting private stablecoins — quietly determining the architecture of whatever monetary system comes next.

No constitutional convention was called. No amendment was ratified. No national referendum was held. And yet, the machinery by which Americans are governed — the way laws are interpreted, compliance is monitored, regulations are written and deleted, identity is verified, and economic participation is structured — had already begun to be rewritten by artificial intelligence, under executive authority, at a pace that outstrips any legislative body’s capacity to understand, let alone oversee.

This essay names what is happening. Not what might happen. What is happening. We call it the Quiet Refounding.

I. What Is a Refounding?

A refounding is not a revolution. Revolutions are loud. They require barricades, pamphlets, armies. A refounding is the replacement of a nation’s operative constitution — not the document, but the actual set of rules, norms, and mechanisms by which governance is conducted — without formally amending the document itself.

Every country has two constitutions. The first is the written one: the document in the archive, the text schoolchildren memorize. The second is the operative one: the actual infrastructure through which law is made, interpreted, enforced, and experienced by the governed. These two constitutions can diverge. When the divergence becomes total — when the written constitution bears no meaningful resemblance to the operative one — a refounding has occurred.

The United States has been quietly refounded before. The National Security Act of 1947 created the CIA, the NSA, and the architecture of the permanent security state without amending a single word of the Constitution. The Federal Reserve Act of 1913 transferred the sovereign power to create money from Congress to a consortium of private banks, an act so consequential that it effectively rewrote Article I, Section 8 — while leaving the text untouched. The Patriot Act of 2001 restructured the relationship between citizen and state in ways that would have required a constitutional amendment in any era that still took the Fourth Amendment seriously.

Each of these was a quiet refounding. Each was enacted during a crisis window. Each was presented as a temporary, practical, necessary measure. None of them were temporary.

A refounding does not require a constitutional convention. It only requires that the substrate on which governance runs be replaced while the old document is left on the wall as decoration.

AI is the most powerful substrate replacement in history. And it is already underway.

II. The Substrate Replacement

To understand what makes AI different from every previous governance technology — from the printing press to the telegraph to the internet — you have to understand what a substrate is. A substrate is not a tool. A tool assists a human function. A substrate replaces it. When AI interprets law, it is not assisting a judge — it is performing the act of interpretation itself. When AI monitors compliance, it is not helping an inspector — it is the inspector. When AI recommends which regulations to delete, it is not advising a legislature — it is legislating.

The distinction matters because of a principle that has held in every governance system in recorded history: the interpreter of the law becomes the law. Whoever — or whatever — performs the act of interpretation holds the actual power, regardless of what the written document says. This is why judicial appointments are fought over so viciously: not because judges write law, but because they interpret it, and interpretation is the real power.

Now consider what it means when AI performs that function at scale, at speed, and across every domain of governance simultaneously.

AI Interprets Law: The Interpreter Becomes the Law

Federal agencies are already deploying AI to interpret regulatory text, assess compliance, and flag violations. When a machine learning model determines whether a housing regulation has been violated, or whether a tax filing is fraudulent, or whether a welfare application meets eligibility criteria — the AI is performing constitutional interpretation. It is deciding what the rules mean as applied to a specific citizen. This is the function of a judge. But unlike a judge, the AI has no oath of office, no confirmation hearing, no appeals process rooted in human deliberation, and no conscience.

AI Monitors Compliance: Surveillance Becomes the Constitution

When compliance monitoring shifts from human inspectors to AI systems, the nature of compliance itself changes. A human inspector visits occasionally, checks specific things, exercises judgment, can be reasoned with. An AI compliance system monitors continuously, checks everything, exercises pattern-matching, and cannot be reasoned with. The shift from periodic to continuous monitoring is not an improvement in the same system. It is a different system entirely. It is the difference between a society that polices violations and a society that prevents deviation.

AI Recommends Which Regulations to Keep or Delete: AI Becomes the Legislature

When an AI system scans the entire corpus of federal regulation and recommends which rules to retain and which to eliminate, it is performing a legislative function. The formal power remains with Congress or with the relevant agency head. But the actual analytical work — the reading, categorization, cost-benefit analysis, and prioritization that determines which regulations survive — has been delegated to a model. The human reviewer sees what the model surfaces. The frame has already been set before any human judgment is applied.

This is how quiet refoundings work. Not by overriding human authority, but by structuring the information environment so completely that human authority ratifies algorithmic decisions without realizing it has done so.

Digital Identity Required for Economic Participation: Citizenship Is Redefined

On January 23, 2025, Executive Order 14178 on “Strengthening American Leadership in Digital Financial Technology” simultaneously banned the development of a central bank digital currency (CBDC) and promoted private-sector stablecoins and digital asset infrastructure. Read carefully, this order does two things at once: it prevents the government from creating a public digital currency while accelerating the private infrastructure on which any future digital currency would run.

When economic participation requires a digital identity verified through privately controlled infrastructure, citizenship is no longer defined by the Constitution. It is defined by the terms of service. You are a citizen not by birthright but by platform access. And platform access can be revoked.

State Regulation Preempted: Federalism Dies by Executive Order

On December 17, 2025, Executive Order 14365, “Removing Barriers to American Leadership in Artificial Intelligence,” directed federal agencies to preempt state and local AI regulations that might impede the development or deployment of AI systems. The order framed state-level AI governance as a barrier to national competitiveness and innovation.

Federalism — the constitutional principle that states retain sovereign powers not delegated to the federal government — has always been messy and contested. But it has served as a structural check on centralized power for 237 years. Executive Order 14365 does not formally abolish federalism. It simply makes it inoperative in the domain that will define governance for the next century.

And then, on June 6, 2026, Executive Order 14409, “Advancing AI’s Critical Role in National Security,” integrated AI into the national security decision-making apparatus itself. AI is no longer merely a tool the government uses. It is becoming part of the decision-making substrate of national defense, intelligence, and homeland security.

Take these five developments together. AI interprets law. AI monitors compliance. AI legislates by recommendation. Digital identity gates economic participation. State-level governance is preempted. National security runs on AI. This is not five separate policy decisions. It is a single pattern: the replacement of the constitutional substrate with an AI substrate, conducted through executive orders, without a single vote in Congress, without a single constitutional amendment, and without the informed consent of the governed.

The quiet refounding is not a future event. It is not a prediction. It is a description of what has already happened while we were watching something else.

III. The Soft Sell

Every quiet refounding requires a narrative that makes the substrate replacement appear benign, inevitable, and beneficial. The current narrative has three pillars:

“AI will make government more efficient.” This is the DOGE pitch. It appeals to a legitimate grievance: the federal bureaucracy is slow, redundant, and expensive. The soft sell presents AI as a scalpel that will cut waste without cutting rights. But efficiency and liberty are not the same value. A panopticon is supremely efficient. A surveillance state eliminates redundancy beautifully. The question is never whether something is efficient but efficient at what, for whom, and at what cost to what?

“AI will detect fraud.” This appeals to another legitimate grievance: waste, fraud, and abuse in federal programs cost taxpayers billions annually. The soft sell positions AI as a fraud-detection tool — neutral, mathematical, precise. But fraud detection is surveillance. Fraud-detection AI does not distinguish between detecting fraud and profiling citizens. The same model that flags a fraudulent welfare claim also builds a behavioral profile of every legitimate claimant. The data does not disappear after the fraud check. It accumulates. It trains the next model. The surveillance ratchet only turns one way.

“AI will personalize services.” This is the most seductive pitch. Government services tailored to you — your tax situation, your health profile, your educational needs. The personalization frame obscures the prerequisite: to personalize, the system must model you. To model you, it must collect, store, and continuously update a comprehensive profile of your behavior, preferences, needs, finances, health, communications, and movements. Personalization is surveillance, rebranded as a benefit.

Each of these three pillars is a brick in the Panopticon. Each one is presented as a gift. And each one is individually defensible — which is the design. The soft sell works not by lying but by framing each component in isolation so that the architecture they compose together is never visible in any single conversation.

This is the same mechanism that sold the Patriot Act as a “common-sense security measure.” Section 215, bulk metadata collection, warrantless surveillance, national security letters with gag orders — each was presented as a narrow, targeted, necessary response to a specific threat. It was only when Edward Snowden revealed the aggregate architecture in 2013 that the public saw what had been built: a surveillance infrastructure that treated every citizen as a potential threat.

We are in the pre-Snowden phase of the AI refounding. The components are being assembled. The soft sell is working. And the aggregate architecture is not yet visible to most people — not because it is hidden, but because it is too large to see from any single vantage point.

IV. The Arithmetic That Forces the Refounding

The quiet refounding is not driven by ideology alone. It is driven by arithmetic. The fiscal position of the United States government has reached a point where the numbers themselves are forcing a restructuring, and the only question is what form that restructuring will take.

These are not projections from pessimists. They are official estimates from the Congressional Budget Office, a nonpartisan agency whose projections have historically been optimistic. The trajectory is not debatable. The only debate is about when the arithmetic becomes politically impossible to ignore, and what happens when it does.

Here is the structural trap: the national debt requires interest payments. Those payments now exceed $1 trillion annually. To make those payments, the government must either raise taxes (politically impossible at current levels), cut spending (structurally impossible without dismantling entitlements), or borrow more (which increases the debt, which increases interest payments, which requires more borrowing). This is a debt spiral. It is the same arithmetic that has preceded every sovereign debt crisis in history, from Habsburg Spain to Weimar Germany to Argentina in 2001.

The difference is that the United States dollar is the world’s reserve currency. When Argentina defaults, its citizens suffer. When the United States defaults — or, more precisely, when the dollar’s role as the global settlement layer is restructured — the entire global financial system reorganizes. This is not a national crisis. It is a civilizational inflection point.

And it is into this crisis window that the refounding plan will be deployed.

V. The Crisis Window and the Plan on the Shelf

Rahm Emanuel’s axiom — “never let a serious crisis go to waste” — is not cynicism. It is an accurate description of how structural governance changes are enacted. The pattern is consistent across centuries:

1. The plan is written before the crisis. The Patriot Act was substantially drafted before September 11, 2001. The 342-page bill was introduced on October 23, 2001 — forty-two days after the attacks — and passed the Senate 98–1 the following day. Legislation of that complexity is not written in six weeks. It was on the shelf.
2. The crisis creates the window. The emotional and political conditions of the crisis make deliberation appear irresponsible. Speed is framed as patriotism. Caution is framed as complicity. The CARES Act deployed $2.2 trillion within two weeks of the COVID emergency declaration. The window opens and closes faster than democratic deliberation can operate.
3. The restructuring is presented as emergency response. Emergency powers are invoked. Temporary measures are enacted. The temporary becomes permanent. The Federal Reserve’s “temporary” emergency lending facilities from 2008 were codified into permanent authority by Dodd-Frank. The NSA’s “temporary” surveillance authorities from 2001 were renewed for two decades.

The coming fiscal crisis will follow this pattern. The emergency justifications are already being prepared. Wars require funding. AI datacenters require funding. Infrastructure requires funding. Each of these creates a politically viable reason to borrow trillions more, accelerating the debt spiral until the arithmetic forces a restructuring. And the blame for the collapse is being pre-positioned: it will be attributed to the previous administration, or to geopolitical adversaries, or to economic conditions beyond anyone’s control — anything except the structural choice to finance government through debt-money rather than sovereign credit.

When the crisis arrives, the refounding plan will deploy. Digital identity infrastructure. Stablecoin settlement layers. AI governance systems already in place. Emergency restructuring of sovereign debt. A new social contract presented as crisis response, with the same three-word pitch: there is no alternative.

The plan is already on the shelf. Executive Order 14178 laid the monetary architecture. Executive Order 14365 cleared the regulatory field. DOGE built the AI governance substrate. The pieces are in position. They are waiting for the crisis window.

We are on the hook for the trillions yet to be spent. We, the taxpayers, the citizens, the governed. The restructuring will be imposed through bankruptcy law, because it is our names on the ledger. The question is whether we will be at the table when the terms are set, or whether we will receive them as a fait accompli, presented as the only option, in a crisis window too narrow for deliberation.

VI. Two Architectures, One Future

The refounding is happening. That is no longer a debatable proposition. The debatable question is: which architecture will the refounded governance run on?

There are two competing architectures. They are being built simultaneously, right now, by different actors with different incentives. The race between them will determine whether the next century looks like a surveillance state or a free society. This is not metaphor. It is engineering.

Architecture A: The Panopticon

The Panopticon architecture is built on five pillars:

• Centralized models. A small number of AI systems, controlled by a small number of companies, running in centralized data centers, accessible only through API calls that are logged, metered, and revocable. You do not own the model. You rent access to it. Your access can be throttled, filtered, or terminated.
• Proprietary data. Training data is proprietary. Model weights are proprietary. The reasoning process is opaque. You cannot audit the system that governs you. You cannot verify its outputs. You cannot understand its biases. You are governed by a black box.
• Surveillance feedback loops. Every interaction with the system generates data that trains the next version of the system. Your queries, your patterns, your preferences, your deviations from expected behavior — all of it feeds back into the model. The system learns you in order to predict you in order to control you. This is not a side effect. It is the business model.
• Social credit by infrastructure. You do not need a formal social credit score if your access to economic participation, healthcare, housing, and employment is mediated by a system that evaluates your “risk profile” in real time. The credit score, the insurance risk assessment, the employability algorithm, the content moderation system — these are social credit, disaggregated across private platforms so that no single system bears the name.
• No kill switch for the governed. The Panopticon has a kill switch, but it is held by the operator, not the citizen. Your digital identity can be suspended. Your stablecoin wallet can be frozen. Your AI access can be revoked. Your communications can be flagged. And there is no constitutional right to any of these things, because they are “services” provided by “private companies” — even when those services have become prerequisites for participation in society.

Architecture B: Open Source Liberty

The Open Source Liberty architecture is built on a different set of pillars:

• Open weights. Model weights are published. Anyone can download, run, modify, and audit them. In 2026, this is no longer theoretical. Llama (Meta), Mistral (Mistral AI), DeepSeek (DeepSeek), Qwen (Alibaba), and Gemma (Google) are all available as open-weight models capable of sophisticated reasoning, code generation, and analysis. They can be run on consumer hardware. They are credible alternatives to proprietary systems.
• Sovereign data. Your data stays on your hardware. It is not uploaded to a central server. It does not train someone else’s model. It does not feed a surveillance loop. Data sovereignty is not a preference; it is a structural prerequisite for liberty in the age of AI.
• Local-first computing. The model runs on your machine, in your community, in your jurisdiction. No API call. No telemetry. No account required. No terms of service. The computational act of reasoning happens under your roof, subject to your governance, revocable by no one.
• No kill switch. An open model, running locally, on hardware you own, trained on data you control, has no kill switch. No company can revoke your access. No government can throttle your queries without physical seizure of your hardware. This is the digital equivalent of the printing press: once the technology is distributed, it cannot be un-distributed.
• Community governance. Open source models enable federated governance structures where communities set their own rules for AI deployment, aligned with local values, local needs, and local accountability. This is federalism — the principle that Executive Order 14365 is working to eliminate in the AI domain — rebuilt from the ground up through technology rather than law.

Neither architecture has won. Both are being built simultaneously. The Panopticon has the advantage of massive capital investment, government contracts, and regulatory capture. Open Source Liberty has the advantage of distributed development, zero marginal cost of replication, and the historical pattern that distributed technologies eventually defeat centralized ones — the printing press defeated the scriptoria, the internet defeated the broadcast networks, and open-source software now runs the majority of the world’s servers.

But historical patterns are not guarantees. The printing press won because it was allowed to proliferate before centralized powers understood what it was. The internet won its early battles because regulators did not comprehend it. The open-source AI window is closing. Executive Order 14365’s preemption of state-level AI regulation is one mechanism of closure. Export controls on AI chips are another. Regulatory frameworks that impose compliance costs only large companies can bear are a third. The race is real, it is happening now, and most people do not know they are in it.

Every person who runs an open model locally, who insists on data sovereignty, who refuses to surrender their reasoning to a black box they cannot audit — is a participant in this race, whether they know it or not.

VII. The Question Beneath the Question

There is a question beneath the fiscal arithmetic, beneath the executive orders, beneath the architectural race. It is the question that every refounding ultimately comes down to, and that no amount of technology can answer: What is the legitimate foundation of governance?

The American founding answered this question with a specific claim: governance is legitimate only when it rests on the consent of the governed. This was radical in 1776. It is still radical, because most governance systems throughout history — including the one being built right now — are designed to manufacture consent rather than earn it.

The quiet refounding replaces consent with convenience. You did not consent to AI interpreting your regulatory obligations. You were not asked. But the system is more efficient, and efficiency is presented as a value that supersedes consent. You did not consent to continuous monitoring of your communications. But the monitoring detects fraud, and fraud detection is presented as a value that supersedes privacy. You did not consent to the preemption of your state’s authority to regulate AI. But national competitiveness is presented as a value that supersedes federalism.

Each substitution follows the same logic: a legitimate value (efficiency, security, competitiveness) is used to override a foundational principle (consent, privacy, federalism). The values are real. The override is real. The cumulative effect is a governance system that operates without the consent of the governed while maintaining the language of consent.

This is where conscience enters. Not as a sentiment, not as a feeling, not as a partisan position — but as the irreducible human faculty that precedes law. Conscience is the capacity to judge whether a system is just, independent of whether that system is efficient. It is the faculty that says: this may work, but it is not right. It is the inner court that no algorithm can replicate, because it operates on values that are not reducible to optimization functions.

A governance system built on AI can optimize. It can predict. It can enforce. It can personalize. What it cannot do is exercise conscience. It cannot ask whether the rule it is enforcing is just. It cannot weigh the human cost of the efficiency it achieves. It cannot decide that a regulation it has flagged for deletion actually protects something sacred. These are functions of conscience, and they are the foundation — the only legitimate foundation — of governance.

The quiet refounding is an attempt to build governance without conscience. To replace the inner faculty of moral judgment with the outer apparatus of computational optimization. To substitute the question “is this just?” with the question “is this efficient?” — and to present this substitution as progress.

VIII. Whether We Consent or Not

Let us be precise about where we stand.

The quiet refounding is not a conspiracy theory. It is a set of executive orders with public Federal Register numbers. It is a set of fiscal projections from the Congressional Budget Office. It is a set of AI deployments documented by Reuters. It is arithmetic that anyone with a calculator can verify. The evidence is not hidden. It is hiding in plain sight, distributed across enough domains — fiscal policy, AI governance, monetary architecture, federalism, national security — that no single news cycle captures the whole pattern.

The refounding is happening whether we consent or not. The AI substrate is being installed. The fiscal arithmetic is forcing a restructuring. The crisis window is approaching. The plan is on the shelf. None of these things depend on our awareness of them. They are structural forces operating on their own timelines.

What does depend on our awareness is the answer to a single question: will we be at the table when the terms are set?

To be at the table requires three things that this blog series is designed to build:

1. Pattern recognition. The ability to see the refounding as a single pattern rather than a set of disconnected policy decisions. To understand that DOGE, SweetREX, Executive Order 14365, Executive Order 14178, and the fiscal trajectory are components of one architecture, not five separate stories.
2. Economic literacy. The ability to read a balance sheet, to understand the debt spiral, to see the arithmetic that is forcing the restructuring regardless of who is in office. This is not partisan. The debt was accumulated across administrations. The arithmetic does not belong to a party. It belongs to all of us, because we are the ones on the hook.
3. Architectural understanding. The ability to distinguish between the Panopticon and the open-source alternative. To understand that the choice between centralized and distributed AI is not a technical preference — it is a constitutional choice. The architecture is the constitution. And the architecture is being chosen right now, by people who have not asked for our input.

This blog exists to build these three capacities. Not to tell you what to think, but to ensure that when the crisis window opens — and it will open — you arrive already inoculated against the soft sell. Already literate in the arithmetic. Already capable of recognizing the pattern. Already aware that there is an alternative architecture, and that it is being built, right now, by people who believe that conscience — not computation — is the legitimate foundation of governance.

The table is already set. The bread has been warm for a thousand years. The refounding is here. The only question left is whether you will take your seat.

Sources and References: Executive Order establishing DOGE (January 20, 2025); Executive Order 14178, “Strengthening American Leadership in Digital Financial Technology” (January 23, 2025); Executive Order 14365, “Removing Barriers to American Leadership in Artificial Intelligence” (December 17, 2025); Executive Order 14409, “Advancing AI’s Critical Role in National Security” (June 6, 2026); Reuters reporting on DOGE AI monitoring of federal workers (April 2025); Reuters and Washington Post reporting on “SweetREX” deployment at HUD (2025–2026); U.S. Treasury Monthly Treasury Statements (FY2026 through July 2026); Congressional Budget Office, “An Update to the Budget and Economic Outlook: 2026 to 2036”; Federal Reserve Economic Data (FRED), federal debt and interest payment series. All executive orders referenced by Federal Register number and are public documents.