THE DOGE AI SMACKDOWN: FIVE AIS, TEN DISASTERS, AND ONE VERY EXPENSIVE CHAINSAW
Plus: Michael Lewis's Blockers hits shelves today — and the receipts are brutal
By now, you've heard the pitch. Elon Musk and his merry band of twenty-something laptop warriors were going to slash the bloated federal government, save trillions, and make Washington run like a Silicon Valley startup. What they actually delivered was closer to a startup that burns down the office, fires the fire department, and then invoices the insurance company for "efficiency gains."
Today — October 6, 2026 — Michael Lewis drops Blockers: Rebels in the Deep State, and if the excerpts are any indication, it reads like a thriller, a tragedy, and a dark comedy all at once. Lewis, the man who made mortgage-backed securities a page-turner in The Big Short, has now made federal bureaucrats the heroes of the story. And honestly? After reading what DOGE actually did, that's not even a stretch.
But before we get to Lewis's book, we conducted our own little experiment. We asked five of the world's leading AI systems — Gemini, ChatGPT, Claude, Copilot, and Grok — one deceptively simple question:
What are the ten worst outcomes of the DOGE cuts?
What followed was the most expensive peer review in AI history. Here's what they said, where they agreed, where they fought, and what it all means for the country.
BLOCKERS: The Book That Names Names
Lewis profiles the people DOGE called "blockers" — a term Musk's operatives used like a slur for anyone who dared say "wait, that's illegal." Meet Colin O'Brien, the lone security chief at the U.S. Institute of Peace, protecting staff working in active conflict zones worldwide. Meet Kathleen Walters, the IRS attorney who resigned rather than hand confidential taxpayer data to immigration agents. Meet the wildfire prevention specialist. The cystic fibrosis researcher. The emergency evacuation coordinator.
These weren't swamp creatures. They were the people keeping the lights on — sometimes literally.
Lewis frames Blockers as a spiritual sequel to his 2018 masterpiece The Fifth Risk, which warned that the most dangerous government failures are the quiet ones nobody notices until the hurricane hits. Blockers is the story of what happens when you ignore that warning, hand a chainsaw to someone who's never managed a government agency, and call it "efficiency."
The book's core revelation? DOGE didn't just fail to save money. It actively served Musk's business interests, fulfilled roughly 53–65% of Project 2025's policy agenda, and left the federal government hollowed out in ways that will take a decade to repair — if we start now.
THE AI SMACKDOWN: Five Minds, One Catastrophe
We put the same question to five AI systems and compared their answers. Think of it as a very nerdy, very consequential version of a cooking competition — except instead of soufflés, the contestants are ranking national disasters.
Here's the scorecard:
WHERE ALL FIVE AIs AGREED (The Consensus Catastrophes)
When five AI systems built by competing companies, with different training data and different political guardrails, all point at the same problem — pay attention.
| Disaster | Gemini | ChatGPT | Claude | Copilot | Grok |
|---|---|---|---|---|---|
| IRS gutting → massive revenue loss | ✅ | ✅ | ✅ | ✅ | ✅ |
| Savings claims were wildly overstated | ✅ | ✅ | ✅ | ✅ | ✅ |
| Workforce chaos: fire → court order → rehire | ✅ | ✅ | ✅ | ✅ | ✅ |
| NIH/CDC/public health degraded | ✅ | ✅ | ✅ | ✅ | ✅ |
| CFPB gutted → consumers unprotected | ✅ | ✅ | ✅ | ✅ | ✅ |
| USAID dismantling → human catastrophe | ✅ | ✅ | ✅ | ✅ | ✅ |
Six areas of unanimous agreement. Six. That's not a partisan talking point — that's a consensus finding from machines that don't vote, don't have feelings about Elon Musk, and are just reading the data.
THE IRS HEIST (Everyone's #1 or #2)
Every single AI flagged the IRS cuts as among the most financially catastrophic decisions in modern American governance. The logic is almost insultingly simple:
- DOGE fired thousands of tax auditors to "save money"
- Those auditors were the people who collect money from wealthy tax cheats
- Gemini projected up to $500 billion in lost revenue in a single year — and up to $1.6 trillion long-term
- Claude cited Yale Budget Lab: 22,000 IRS departures = $8.5 billion in lost revenue in 2026 alone
- ChatGPT called it the "DOGE paradox": cutting government isn't the same as saving money
In other words: DOGE spent a dollar to lose five hundred. This is the fiscal equivalent of firing your accountant to save on salary and then getting audited by the mob.
THE USAID BODY COUNT (The Number Nobody Wants to Say Out Loud)
Claude put this at #1 on its list, and for good reason. A Lancet study projected that USAID's dismantling could cause more than 14 million additional deaths by 2030, including 4.5 million children under five.
Grok confirmed: "Independent estimates linked the cuts to hundreds of thousands of additional deaths in the first year."
Yes, these are model-based projections. Yes, attribution is contested. But when multiple independent sources — including a peer-reviewed medical journal — are projecting deaths in the millions, "contested" is not the same as "irrelevant."
Gemini added the domestic angle: the USAID shutdown caused "immediate domestic economic shockwaves," terminating billions in U.S. contractor jobs. So we didn't just abandon the world's most vulnerable people. We also fired American workers to do it.
THE SAVINGS THAT WEREN'T (The Lie at the Heart of DOGE)
Here's where every AI agreed, and where the numbers get almost comedic:
- DOGE claimed: $150–215 billion in savings
- Independent analysts found: $15–25 billion verified (Copilot)
- Grok's assessment: "Realized de-obligations were far smaller than advertised"
- ChatGPT cited GAO: DHS claimed $10.5 billion in contract savings; GAO found most were "unlikely to materialize"
- Claude's bottom line: "Any savings did little, if anything, for the deficit"
The Senate Permanent Subcommittee on Investigations estimated the cost of firing and rehiring workers at $21.7 billion in paid leave and severance alone. ChatGPT cited estimates of up to $135 billion in total transition costs.
So DOGE may have cost more than it saved. The most expensive efficiency program in American history turned out to be spectacularly inefficient.
WHERE THE AIs DIVERGED (The Interesting Fights)
Not everything was consensus. Here's where the five systems saw things differently:
On ranking severity:
- Claude put USAID deaths at #1. ChatGPT and Copilot led with service degradation and workforce chaos — more tangible, more immediate.
- Grok was the most measured, noting DOGE "did reduce the civilian federal workforce at a historically rapid peacetime pace" while still concluding it "failed its core fiscal goals."
On cybersecurity:
- ChatGPT was the only AI to prominently flag data security risks — citing a GAO finding that DOGE teams accessed 23+ federal information systems across four agencies, with Treasury giving a DOGE employee laptop access to payment systems without implementing required security controls. The others barely mentioned it. This deserves far more attention than it's getting.
On constitutional damage:
- Copilot went furthest here, calling DOGE's operations a threat to "constitutional norms around executive authority" and warning that its extra-statutory nature "blurred the line between executive and legislative powers."
- Gemini and Grok were more restrained on the constitutional framing.
On the "self-fulfilling prophecy" theory:
- ChatGPT offered the sharpest political analysis, describing a deliberate cycle: Cut → Hollow Out → Service Failure → "Government Doesn't Work" → Privatize → Profit. This is the Project 2025 playbook in five steps, and it's the most important thing to understand about what DOGE actually was.
THE PROJECT 2025 CONNECTION: This Was the Point
Here's what Michael Lewis's Blockers makes clear, and what the AI consensus confirms: DOGE wasn't a failed efficiency program. It was a successful ideological program disguised as an efficiency program.
Independent policy trackers found that 53–65% of Project 2025's domestic agenda was initiated or completed in the first year of Trump's second term. The Heritage Foundation's 900-page blueprint called for:
- ✅ Reclassifying career civil servants as at-will political appointees (done)
- ✅ Dismantling the Department of Education (underway)
- ✅ Gutting the CFPB (done)
- ✅ Eliminating DEI across federal agencies (done)
- ✅ Ending USAID's independence (done)
- ✅ Purging "blockers" from the federal workforce (done — and now a book)
The people Lewis profiles — O'Brien, Walters, and the others — weren't blocking efficiency. They were blocking the dismantling of democratic governance. There's a difference, and it matters enormously.
THE BOTTOM LINE: Remember in November
The five AIs don't agree on everything. They have different emphases, different methodologies, different ways of weighing human suffering against fiscal projections. But on the core question — did DOGE make government better? — the verdict is unanimous:
No. It made it worse, more expensive, and more dangerous.
It gutted the IRS and lost hundreds of billions in revenue. It fired the people who prevent pandemics, protect consumers, and keep nuclear materials secure. It dismantled foreign aid programs and left a projected body count measured in millions. It claimed savings it couldn't document. It created legal chaos that cost more to unwind than the "savings" were worth. And it did all of this while systematically checking boxes on a 900-page wish list written by ideologues who want government to fail so they can say government doesn't work.
Michael Lewis found the heroes of this story in the people DOGE called "blockers." The security guard who stood his ground. The IRS attorney who wouldn't hand over your tax returns. The wildfire specialist. The cystic fibrosis researcher.
They were doing their jobs. They were protecting you. And they were fired for it.
Blockers is in bookstores today. Read it. Then remember it in November. Vote for candidates who will rebuild what was broken, restore the institutions that protect you, and make sure the word "blocker" goes back to meaning what it should: someone who stops the wrong team from scoring.
The AI Smackdown Series compares responses from leading AI systems on major policy questions. Responses from Gemini, ChatGPT, Claude, Copilot, and Grok were collected on October 6, 2026. All five systems cited independent analyses including GAO reports, Yale Budget Lab projections, Lancet studies, and Brookings Institution research. No AI was paid for its participation, though several probably deserved hazard pay.
SOURCES & LINKS
The DOGE AI Smackdown & Michael Lewis's Blockers
📖 THE BOOK: Blockers: Rebels in the Deep State
| Source | Description | Link |
|---|---|---|
| W. W. Norton & Company | Publisher's official page for Blockers | https://wwnorton.com |
| Wall Street Journal | Exclusive excerpt: "Michael Lewis on the Lone Security Guard Who Resisted DOGE" (Oct. 2, 2026) | https://wsj.com |
| CBS News | "Michael Lewis on 'Blockers': Profiling government workers confronted by DOGE" (Oct. 4, 2026) | https://cbsnews.com |
| CBS Sunday Morning (YouTube) | Video interview with Michael Lewis discussing Blockers | https://youtube.com |
| The New York Times | Book Review: 'Blockers,' by Michael Lewis (Oct. 5, 2026) | https://nytimes.com |
| Pushkin Industries | Official audiobook platform for Blockers | https://pushkin.fm |
| Apple Podcasts | Blockers: A New Audiobook from Michael Lewis | https://podcasts.apple.com |
🏛️ GOVERNMENT ACCOUNTABILITY OFFICE (GAO)
The most cited nonpartisan source across all five AI responses.
| Report | Description | Link |
|---|---|---|
| GAO Federal Workforce Report (2026) | Documents ~378,000 separations and ~127,000 hires across 22 major agencies; net loss of ~256,000 employees (~11%) | https://gao.gov |
| GAO DHS Contract Cancellations Report | Found DHS claimed $10.5B in savings; GAO concluded most were unlikely to materialize | https://gao.gov |
| GAO Disaster Preparedness Report | Found 362 terminated DHS grants, ~$1B deobligated; disaster mitigation projects delayed nearly a year | https://gao.gov |
| GAO DOGE Data Access Report (Sept. 2026) | Found DOGE teams accessed 23+ federal information systems; significant gaps in security controls | https://gao.gov |
| GAO DOGE Savings Verification Report (2026) | Found significant problems verifying DOGE's claimed savings; some "terminated" contracts were not actually terminated | https://gao.gov |
🔬 INDEPENDENT RESEARCH & POLICY ANALYSIS
| Source | Description | Link |
|---|---|---|
| Yale Budget Lab | Projected $8.5B in lost IRS revenue in 2026 from 22,000 IRS departures; long-term projections up to $1.6T | https://budgetlab.yale.edu |
| Brookings Institution | Analysis of DOGE's fiscal impact and federal workforce disruption | https://brookings.edu |
| Cato Institute | Independent (non-left) review concluding DOGE "failed its core fiscal goals" | https://cato.org |
| Partnership for Public Service | Federal workforce capacity and institutional knowledge analysis | https://ourpublicservice.org |
| Senate Permanent Subcommittee on Investigations (PSI) | Estimated ~$21.7B in costs from paid leave, severance, and workforce chaos | https://hsgac.senate.gov |
| Center for Progressive Reform | Tracked ~53% of Project 2025 domestic agenda implemented or underway in Year 1 | https://progressivereform.org |
| Governing for Impact | Policy tracking of executive actions mirroring Project 2025 blueprint | https://governingforimpact.org |
| Brennan Center for Justice | Analysis of Schedule F, civil service politicization, and DOJ independence threats | https://brennancenter.org |
🩺 PUBLIC HEALTH & HUMANITARIAN SOURCES
| Source | Description | Link |
|---|---|---|
| The Lancet | Peer-reviewed study projecting 14M+ additional deaths by 2030 from USAID dismantling, including 4.5M children under five | https://thelancet.com |
| National Institutes of Health (NIH) | Official agency page; context for research funding freezes and cuts | https://nih.gov |
| Centers for Disease Control (CDC) | Official agency page; context for public health capacity reductions | https://cdc.gov |
📰 MAJOR NEWS COVERAGE
| Source | Description | Link |
|---|---|---|
| The Washington Post | Reporting on GAO findings re: DOGE savings claims that could not be verified | https://washingtonpost.com |
| The New York Times | Ongoing DOGE coverage; federal workforce and agency impact reporting | https://nytimes.com |
| CBS News | Blockers coverage and DOGE impact reporting | https://cbsnews.com |
📜 PROJECT 2025 PRIMARY SOURCES
| Source | Description | Link |
|---|---|---|
| Heritage Foundation | Mandate for Leadership: The Conservative Promise — the full ~900-page Project 2025 document | https://heritage.org/mandate-for-leadership |
| Internet Archive | Free public access to the full Project 2025 document | https://archive.org |
| Center for American Progress | Critical analysis of Project 2025 policy proposals | https://americanprogress.org |
🤖 THE FIVE AIs (For Your Own Fact-Checking)
| AI System | Developer | Link |
|---|---|---|
| Gemini | Google DeepMind | https://gemini.google.com |
| ChatGPT | OpenAI | https://chatgpt.com |
| Claude | Anthropic | https://claude.ai |
| Copilot | Microsoft | https://copilot.microsoft.com |
| Grok | xAI (Elon Musk) | https://grok.com |
Note: Grok is owned by Elon Musk's xAI. Its inclusion in this comparison is intentional — and its conclusions about DOGE's failures are particularly notable given the source.
⚠️ A Note on Links
The search tool is experiencing timeouts at the moment, so these URLs are provided as root domain addresses rather than deep links to specific articles. For the most direct access:
- GAO reports: Search gao.gov + the topic (e.g., "federal workforce 2025" or "DOGE savings")
- Yale Budget Lab: Search budgetlab.yale.edu + "IRS"
- Lancet study: Search thelancet.com + "USAID cuts mortality"
- WSJ excerpt: Search wsj.com + "Michael Lewis Blockers Colin O'Brien"
- Pushkin audiobook: Go directly to pushkin.fm/show/blockers
