Sabitlenmiş Tweet
The Anthropic-Pentagon Fight and the Fourth Institutional Cycle
George Friedman argues in The Storm Before the Calm that American history runs on ~80-year institutional cycles. Each one doesn't just redistribute power — it redefines the basis of legitimacy for federal authority.
The Founding grounded legitimacy in consent of the governed. The Civil War elevated the federal government above the states. The post-WW2 settlement expanded federal power into daily life, grounding legitimacy in technocratic expertise — the idea that credentialed professionals, guided by data, should manage the economy, defense, healthcare, and education.
We are now in the fourth crisis. The technocratic model is failing — not because expertise is wrong, but because the expert class became self-referential. Experts validate each other, credentialing institutions gatekeep who counts, and the federal apparatus answers primarily to its own internal logic rather than to democratic input.
The evidence is everywhere. California's high-speed rail: $18 billion spent, zero trains running, $765 million consumed by environmental review alone, 70% of the workforce consultants rather than builders. The NEVI charging program: $7.5 billion allocated, 384 ports deployed after four years while Europe and China built comparable networks for a fraction of the cost. The federal government spends roughly $23.5 billion per year on management consulting. The process doesn't just slow things down — it has become the product.
Left-populists call it capture by corporate interests. Right-populists call it the deep state. Both are describing the same structural problem.
Enter the Anthropic-Pentagon standoff.
In February 2026, Anthropic baked its acceptable use policy — including prohibitions on mass surveillance and autonomous weapons — directly into its DoD contract. The Pentagon's position: legality of use is the government's call, not the vendor's. Trump directed federal agencies to cease using Anthropic's technology.
The substance is almost beside the point. The structural question is who decides. A private company founded by AI safety experts is imposing constraints on the elected executive's control of national defense, via contract terms enforcing what Congress hasn't legislated. The Pentagon's response — "all lawful purposes" — is a reassertion of democratic authority over expert authority.
This maps precisely onto Friedman's framework. The expert class asserting institutional prerogatives over elected democratic authority, and the executive branch pushing back. The counter-move fits too: 430+ employees at Google and OpenAI signed a letter backing Anthropic — expert class solidarity across corporate lines, the technocratic network circling the wagons exactly as the cycle turns against them.
What makes this cycle genuinely different from the prior three is that the technology at the center is itself a form of institutional power.
AI isn't like steel, oil, or nuclear weapons. AI is a decision-making substrate. Whoever controls how AI systems are trained, what values they embed, and what they refuse to do is controlling a new form of governance. Anthropic isn't just a defense contractor with safety concerns — it's an organization that has embedded its values into a cognitive tool that millions rely on. That's a form of power that didn't exist in any prior cycle.
The values encoded into AI systems are currently being decided by a small technocratic class at private companies, with no democratic input, no transparency, and no accountability mechanism.
Three plausible endgames:
The Executive Bargain — the president gets substantially more direct control over defense and intelligence, areas where democratic accountability is clearest. Domestic governance gets devolved. The federal government shrinks but sharpens.
The Platform Model — the federal government transforms from operator to platform. AI becomes the mechanism: instead of 87,000 IRS employees interpreting tax law, AI systems apply rules uniformly. The irony: the technology the government is fighting to control may be the thing that makes its own downsizing possible.
The Negotiated Retreat — historically most consistent. In every prior cycle, the existing order didn't get demolished, it got subordinated to a new organizing principle. The technocratic class retains influence but loses its autonomy, becoming explicitly subordinate to democratic authority. Companies like Anthropic would still advocate for safety, but the legal basis for overriding executive authority on national security would be eliminated. The experts advise; the elected officials decide.
The missing piece: if the platform model simply replaces federal agencies with code — AI handling operational governance, blockchain handling financial infrastructure — and the same technocratic class writes the standards, nothing has changed. You've just made the expert class more efficient at the same unaccountable governance.
The resolution only works if these systems are transparent and publicly auditable. This is where blockchain may play a supporting role — credibly neutral ledgers where rules are encoded, visible, and enforced algorithmically rather than by intermediaries. Not replacing governance, but making it legible.
What's missing is the democratic legitimacy layer that connects AI's operational power and blockchain's transparency to the will of the governed. The AI people have the governance tool. Blockchain offers the transparency ledger. Nobody has designed the constitutional bridge yet.
English
