THE REVIEW
“Where governance meets its own reflection—and winces”
Welcome to the Machine
Dear reader, what you hold before you is a dispatch from the land of unintended consequences—a territory so vast it stretches from government file cabinets stuffed with unanswered requests to poultry farms where farmers hide sick chickens from the officials sent to help them.
This edition examines a peculiar species of failure: systems that defeat themselves while everyone follows the rules. No villains required. No conspiracy needed. Just incentives, procedures, and the inexorable logic of people doing exactly what we asked—producing the opposite of what we wanted.
In these pages you will meet:
- The transparency law that creates opacity (222,000 unanswered requests)
- The efficiency initiative that increased spending (+$248 billion)
- The budget cop whose gun was taken away ($430 billion blocked, no remedy)
- The robot doctor war (machines denying, machines appealing)
- The farmers hiding sick birds (from the aid program meant to help them)
- The clean energy graveyard (2,600 gigawatts going nowhere)
None of this is accidental. It is structural.
What connects these stories isn’t partisan politics. It’s a pattern: rules create compliance costs, costs create workarounds, workarounds defeat the rules’ purpose, and dysfunction becomes the stable state.
“It’s like being in a room and the wall, the roof, and the floor are all targets. You see, because you can’t miss.”
— Elon Musk, on DOGE’s targeting methodology
The machine eats itself, and then it keeps running.
❧ ❧ ❧
“Good Luck With That”
How the Freedom of Information Act became an information cemetery
When CNN submitted a Freedom of Information Act request to the Office of Personnel Management asking about security clearances for DOGE staff, an OPM email responded with three words: “Good luck with that.”
The staff responsible for processing FOIA requests had been fired.
That reply captures American transparency law in 2026. The Freedom of Information Act still exists. The twenty-day response mandate remains on the books. The principle that citizens can see government records hasn’t been repealed. But the machinery is gone, and the people who ran it are gone.
The federal FOIA backlog now exceeds 222,000 requests—a record. DHS alone received 911,000 requests last year. An FBI request under fifty pages takes four months. Over fifty pages? Two years.
Information has a shelf life. A document about a controversial policy is explosive before the policy is enacted. Three years later, it’s a historical footnote. By the time documents arrive, the news cycle has moved on, the officials have left office, and the accountability moment has passed.
“Reporters, as is the American citizen, are in for a boatload of trouble in the next four years.”
— Terry Mutchler, National Freedom of Information Coalition
The dysfunction runs deeper than budget cuts. Agencies have little incentive to respond quickly. The cost of delay—a potential lawsuit years down the line—pales against the immediate political risk of embarrassing disclosure. There’s no prize for transparency, only punishment for releasing the wrong document.
The result: a two-tiered system. Well-funded media organizations sue, triggering court-ordered deadlines. Everyone else waits in a queue that moves at the speed of bureaucratic indifference.
FOIA was designed as a democratic accountability engine. It has become a waiting room where accountability goes to die.
For Further Reading: Perspectives
| Source | Summary | |
|---|---|---|
| PRO | Center for Election Innovation & Research (2025) | “States Protect Election Staff from FOIA Abuse” — 13 states enact laws shielding election workers from frivolous requests |
| CON | Just Security (Nov. 2025) | “FOIA and Deteriorating Federal Transparency Infrastructure” — Documents fee waiver denials, premature closures, institutional decay |
❧ ❧ ❧
The $248 Billion Efficiency
271,000 jobs cut. Spending went up anyway.
The Department of Government Efficiency promised to cut 1 trillion. Then $150 billion. Then it quietly disbanded.
What it actually delivered: $248 billion more in federal spending compared to the prior year.
“An observer who did not know when DOGE started could not identify it in the spending curve,” the Cato Institute reported in December.
DOGE did achieve something. It eliminated 271,000 federal jobs in under ten months—the largest peacetime workforce reduction in American history. Headlines proclaimed victory over bureaucratic bloat.
But here’s the arithmetic that guaranteed failure: roughly 70 percent of federal spending flows through mandatory programs—Social Security, Medicare, Medicaid, debt interest. These outlays operate on statutory autopilot. Firing an EPA analyst does nothing to reduce the cost of a Medicare hip replacement.
A 10 percent workforce cut saves about 7.6 trillion budget.
Worse, the cuts created new costs. The Social Security Administration shed 7,000 workers and now faces millions of pending cases. The IRS lost refund processors. NPR reports hundreds of fired employees are being offered their jobs back. Agencies pay overtime to restore basic functions.
“As I look back at what has been done, I would not want to do it again.”
— Elon Musk, December 2025
The strongest defense comes from Vice President Vance: DOGE matters less for saving money than for “making the bureaucracy responsive to the elected president.” By that measure, perhaps it succeeded. But judged on its stated terms—slashing wasteful expenditures—it is a documented failure.
A New York Times analysis found DOGE’s 13 largest claimed cuts were all incorrect. The two biggest entries—Defense contracts listed as “terminations”—were still active.
The tool could not reach the target.
For Further Reading: Perspectives
| Source | Summary | |
|---|---|---|
| PRO | Washington Times (Jan. 2026) | “DOGE at One Year” — Credits DOGE with eliminating USAID authority and sparking nationwide streamlining |
| CON | CREW (June 2025) | “DOGE’s Big Illusion” — Estimates cuts cost 26B to taxpayers |
❧ ❧ ❧
The Watchdog Who Lost His Teeth
How to kill a law without repealing it
Here is how a statute dies while remaining on the books.
The Impoundment Control Act of 1974 says the president can propose rescinding funds, but if Congress doesn’t agree within 45 days, the money must be spent. For five decades, this worked. Nixon tried to refuse spending; Congress passed a law to stop him. GAO monitored compliance. The system held.
In 2025, that machine broke.
GAO documented six violations: DOT withholding electric-vehicle charger funding. The Institute of Museum and Library Services freezing 81 percent of its budget. FEMA delaying emergency shelter funds. NIH grants paused mid-research.
Total funding blocked: approximately $430 billion.
OMB Director Russ Vought’s response? “Non-events with no consequence. Rearview mirror stuff.”
Then came the killing blow. The DC Circuit ruled that only GAO—not states, not universities, not patients whose cancer trials were paused—has legal standing to sue over impoundment violations.
The House Appropriations Committee then voted to cut GAO’s budget by 49 percent and prohibited the agency from filing impoundment lawsuits without a concurrent resolution for each case.
“The same guy who says he wants to get people off the streets is blocking funding for communities to tackle homelessness.”
— Senator Patty Murray
The feedback loop is complete: the judiciary ruled that only GAO can enforce the law, then the legislature ensured GAO can’t afford to.
The statute remains on the books. Its text hasn’t changed. But the system designed to enforce it has been defunded.
When you defund the referee, the rules become suggestions.
For Further Reading: Perspectives
| Source | Summary | |
|---|---|---|
| PRO | Heritage Foundation (2025) | “Executive Authority and the Power of the Purse” — Presidents possess inherent constitutional authority over spending execution |
| CON | House Appropriations Democrats (Feb. 2025) | “Background on Unlawful Impoundment” — Legal consensus against presidential impoundment; documents specific violations |
❧ ❧ ❧
The Robot Doctor War
When algorithms fight algorithms, patients lose
Somewhere in the servers of a major health insurer, an algorithm is rejecting your claim before a human has read it. Somewhere in your hospital’s billing department, a counter-algorithm is drafting an appeal with more legal citations than your doctor could produce in a week.
Neither machine cares about your diagnosis. Both optimize for throughput.
Welcome to the healthcare AI arms race.
According to the AMA’s 2025 survey, three in five physicians believe insurers’ AI is increasing denials. A class action lawsuit against UnitedHealthcare alleged its algorithm wrongfully denied claims—with approximately 90 percent of denials overturned on appeal.
If 90 percent of denials are overturned, they’re not serving a quality-control function. They’re serving a delay function.
The insurer’s bet: most patients won’t appeal. The data proves them right. Fewer than 1 percent of denied claims are appealed—yet 40 to 90 percent succeed when patients fight back.
Now patients are arming themselves. Counterforce Health built an AI to help patients appeal. Sheer Health auto-generates legal citations. The arms race escalates.
“We’re in an AI arms race where as consumers become more savvy… the insurers will just up the ante on their side.”
— Jennifer Oliva, Indiana University Law School
Beginning January 2026, CMS launched the WISeR model—AI-powered prior authorization for Medicare in six states. Vendors get paid a percentage of money saved by preventing “unnecessary” care.
“The more you deny, the more you get reimbursed,” one consultant observed. “A third grader could see how that’s a problem.”
The deeper issue: healthcare adjudication was designed for a world where generating disputes required human effort. When filing costs drop to zero, the system drowns. A plea for life-saving care becomes indistinguishable from a bot-generated appeal in a stack of 50,000 documents.
This is a denial-of-service attack on the arbitration system itself.
For Further Reading: Perspectives
| Source | Summary | |
|---|---|---|
| PRO | Medical Economics (Oct. 2025) | “AI Could Streamline Prior Authorization” — Health systems report 95%+ first-pass approval rates using AI documentation tools |
| CON | National Health Law Program (Dec. 2025) | “Federal AI Policy Threatens Reform” — AI-driven review creates barriers between patients and necessary care |
❧ ❧ ❧
The Farmers Hiding Sick Chickens
When helping hurts
One Illinois family lost its entire 3,000-hen farm to bird flu in early 2025. Because most of the flock was dead by the time USDA arrived, the family reportedly received only 100,000 they expected.
This isn’t a story about stingy government. It’s a story about incentive design.
Since 2022, highly pathogenic avian influenza has killed over 185 million birds. USDA response costs exceed $1.4 billion, mostly in indemnity payments. To prevent moral hazard—why maintain biosecurity if the government bails you out?—USDA changed the rules.
Now, farms affected by bird flu must complete a biosecurity audit before restocking. Fail the audit? No indemnity if the virus returns.
The policy goal is clear: reward compliance, penalize negligence.
The policy effect may be the opposite.
By tying compensation to passing an audit, the policy creates an economic death penalty for transparency. Farmers who suspect infection face a gamble: invite scrutiny that could disqualify them from aid, or hide early signs and hope it passes.
Bird flu can kill 100 percent of a flock within 48 hours. By the time a farmer knows something’s wrong, the economic damage is done. If reporting triggers an audit they might fail, the rational choice is to minimize transparency.
The result: a surveillance network that receives data from compliant, low-risk actors while high-risk vectors—where the virus is actually spreading—go dark.
“The USDA has built a system that pays for ‘hygiene theatre’ but penalizes ‘disease intelligence.‘”
— Ellen Carlin & Gwendolen Reyes-Illg, STAT News
Sixty-seven commercial poultry premises have had multiple infections. Either biosecurity compliance isn’t working, or farms are hiding problems.
When the goal is intelligence rather than punishment, the system must make disclosure safe.
For Further Reading: Perspectives
| Source | Summary | |
|---|---|---|
| PRO | USDA/NASDA (Feb. 2025) | “Strong Support for USDA’s Actions” — $1B five-pronged strategy with biosecurity investments and updated indemnity |
| CON | STAT News (Jan. 2026) | “We Know How to Prevent Bird Flu” — Current approach is “unsustainable,” creates perverse transparency incentives |
❧ ❧ ❧
The Clean Energy Graveyard
2,600 gigawatts going nowhere
If you could flip a switch and connect every energy project waiting in America’s interconnection queue, you’d have 2,600 gigawatts of generation and storage—more than twice the country’s installed capacity.
You can’t flip that switch.
Historical data show that only 19 percent of projects entering U.S. queues ever reached commercial operation. Eighty-one percent die waiting.
Median time from request to operation now averages five years, up from under two decades ago. The queue isn’t a waiting room. It’s a graveyard.
In 2023, FERC issued Order No. 2023 to fix the backlog, implementing “first-ready, first-served” with stricter milestones. Flush out speculators by requiring real financial commitment.
The logic was sound. The execution was catastrophic.
By overcorrecting for speculation, the policy imposed milestones so stringent that even viable, well-funded projects fail on procedural grounds. A developer puts down a massive deposit. A local permit is delayed 30 days by a town council recess. Under the rules, they’re ejected and lose their deposit.
The Southwest Power Pool requested a waiver to delay opening its 2025 queue window—because the rules are so rigid that the grid operator needed permission to stop following them in order to function.
“Developers now spend more time pleading for waivers at FERC than building projects.”
The reform aimed at efficiency generated a new bureaucracy managing exceptions to its own rigidity.
Meanwhile, fossil fuel plants stay online because the transmission lines that would replace them can’t escape the queue.
The energy transition isn’t blocked by lack of investment or technology. It’s blocked by a process that treats every deadline as terminal.
For Further Reading: Perspectives
| Source | Summary | |
|---|---|---|
| PRO | FERC Fact Sheet (2023) | “Order 2023: Reforming Interconnection” — Reforms reduce speculation by requiring financial commitment |
| CON | Lawrence Berkeley Lab (2025) | “Queued Up” — Documents queue growth from <500 GW to 2,600 GW; median times stretching to five years |
❧ ❧ ❧
EDITORIAL
The Machine That Eats Itself
There is a class of failure more insidious than incompetence or corruption: systems that defeat themselves while everyone follows the rules.
The stories in this edition share a pattern. A law establishes procedures to achieve a legitimate purpose. Procedures create compliance costs. Rational actors find the path of least resistance. That path undermines the original purpose. The system reaches stable equilibrium: rules in force, procedures followed, purpose defeated.
This requires no bad actors. No conspiracy. Only people responding to incentives—which they reliably do.
You cannot fire your way out, because the rule-followers aren’t the problem; the rules are. You cannot investigate your way out, because there’s no conspiracy to uncover; only structures to diagnose. You cannot even legislate your way out, because new laws generate new procedures, new compliance costs, new workarounds.
The only escape is understanding the architecture deeply enough to redesign the incentives.
What would that look like?
Mandatory retrospective accounting. Initiatives promising savings must produce audited reports comparing projections against actual outcomes—including downstream costs. Make symbolic cuts impossible to reward.
Self-executing enforcement. Oversight mechanisms shouldn’t depend on annual appropriations. GAO’s authority should be insulated from the pressures it exists to check.
Symmetric friction. When AI eliminates filing costs, reintroduce economic friction through bonds forfeited for meritless filings. Force algorithms to optimize for accuracy, not volume.
No-fault reporting harbors. When the goal is intelligence rather than punishment, make disclosure safe. In a crisis, learning where problems are beats punishing those who reveal them.
Flexible optionality over pass-fail. Systems that treat every deadline as terminal create waiver chaos. Buffer statuses maintain pressure while accommodating legitimate variation.
These systems aren’t random malfunctions. They’re stable equilibria of poorly designed incentives. They will persist until we redesign the incentives—or until the contradictions become intolerable.
The machine eats itself.
The question is whether we’ll build a better one.
For Further Reading: Perspectives
| Source | Summary | |
|---|---|---|
| PRO | OECD (Apr. 2025) | “Regulatory Policy Outlook 2025” — Framework for understanding feedback loops and evidence-based evaluation |
| CON | PNAS (Jan. 2026) | “Feedback Loops in Complex Systems” — Single-variable optimization destroys system redundancy |
❧ ❧ ❧
Production Note
This edition of The Review was produced through collaboration between human editorial judgment and AI research assistance. Facts have been verified against primary sources: GAO reports, Treasury statements, agency data, court filings, and peer-reviewed analyses. Where claims conflict, we note the conflict. Where evidence is uncertain, we say so.
Your skepticism remains appropriate and encouraged.
Coming Next Week: The Paperwork Fortress—examining why environmental review documents now exceed 5,000 pages and what happens when process becomes purpose. Also: the two-track NEPA system where those who pay get faster review.
© 2026 The Review. All rights reserved.
Editor: Daniel Markham
Research Period: December 2025 – February 2026
Published: Friday, February 7, 2026
The Review is not available to the general public.
Subscription by personal referral only.
If you know, you know.