All Savings Reverted: Inside the GAO's Takedown of DOGE's $110.3B Claim

By Isabella Wilson, Crypto Security Audit Partner The Department of Government Efficiency claimed $17 billion in savings from Defense Health Agency technology contracts. The GAO checked. The contracts had never been modified. Not a seat renegotiated. Not a license revoked. The $17 billion existed as a line item on a public website called the "Receipt Wall." A receipt with no underlying transaction. I read the reverts before the headlines. This time, the GAO beat me to it. On August 6, 2025, the Government Accountability Office published its review of DOGE's savings claims. The total claimed through early July 2025: $110.3 billion. The composition: $61 billion in contract terminations, $49.2 billion in grant cancellations, $113 million in lease reductions. The audit's verdict was clinical: significant data inaccuracies, missing supporting evidence, and claims that could not be verified at scale. I spent the first months of 2023 tracing Alameda's ETH and BTC flows after FTX imploded. The balance sheets said one thing; the chain said another. The GAO just ran that same exercise on the Musk efficiency department, with better spreadsheets and far worse optics. The texture is familiar: a dashboard that looks precise until you check the identifiers underneath. If this were Ethereum, those transactions never landed. They were broadcast, not confirmed. The timing matters here. The GAO report dropped on August 6, 2025. I am writing this from February 2026, which gives me the advantage of watching the market and political reactions settle into their default positions. The White House did not flinch. The contractors are still recalibrating. And the next efficiency program is being designed right now, with the GAO report as its ceiling. Let me clarify a naming matter before the comment section fills up. DOGE here is the Department of Government Efficiency. Not the memecoin. The irony is worth one dry sentence: of the two assets named DOGE, the token has proven easier to verify than the department. A memecoin at least publishes a supply schedule. The department was created by executive order on January 20, 2025, the first day of the second Trump term. Elon Musk, a private citizen with no Senate confirmation and no statutory portfolio, was placed at its head. The Receipt Wall went live on February 17, 2025, twenty-eight days after the order. It published claimed savings item by item, in real time, styled as a public ledger. This mattered beyond the theater. The federal government entered 2025 with roughly $36 trillion in debt. Fiscal sustainability had stopped being a think-tank footnote and had become the central political battleground of the second Trump term. DOGE was the executive branch's answer: cut waste without touching entitlements, cut contracts without cutting services, and prove it all on a website. The Receipt Wall was the proof-of-work. And the enterprise was engineered to be audit-proof — not cryptographically, but procedurally. Created by executive order, it bypassed Congressional appropriations. Staffed by engineers and political operatives rather than career civil servants, it had no institutional memory. It was formally terminated on July 4, 2025, roughly eight months before its projected sunset — and before the GAO's review was complete. Silence is just uncompiled potential energy. DOGE ended before the auditor finished reading its logs. The early termination reads two ways. Defenders call it mission completion. Critics call it an exit before the receipts could be contested. From an auditor's seat, an entity that self-terminates before the audit cycle completes has effectively pre-empted the discovery phase. You do not need to guess which of those readings I assign to it. Start with the numbers, because that is where every meaningful claim either dies or survives. Slice the Receipt Wall into its three ledgers. Contracts: $61 billion claimed. The GAO found 13,476 contracts marked as terminated. More than a quarter of those records lacked identifying details — no contract number, no agency, no vendor. Those are unreviewable by construction. Of the remaining records, only 43% could be tied to contracts that were actually fully or partially terminated. Meaning: at least half of the marked terminations were misclassification, miscounting, or pure invention. Even granting full credit for the verified 43%, the contract claim drops from $61 billion to roughly $26 billion — and that figure has not yet been discounted for the unidentifiable quarter sitting underneath it. There is an auditor's rule that governs this: a claim that cannot be reconstructed is treated as zero until evidence appears. That rule is not paranoia. It is the discipline that separates a finding from a headline. Grants: $49.2 billion claimed. The GAO found that 96% of the figure lacked sufficient information to verify the calculation method. Not disputed. Not partially substantiated. Unverifiable. This is the largest non-contract category, and it is a black box. If a DeFi protocol told me its yield was computed from a vault that I could not inspect, I would not call that yield anything but an assertion. Leases: $113 million claimed. Actual verified savings: $31.8 million. That is 28% of the claim. Of the 264 leases counted, 108 were already being wound down before the executive order creating DOGE was signed. The department took credit for depreciation that was already in motion. And then there is the exemplar, the line item I keep returning to. The Defense Health Agency. DOGE claimed seventeen billion dollars in savings from technology contracts supporting roughly 700 military healthcare facilities. The GAO examined the claim. The contracts were never modified. The claimed savings were attributed to work that never happened — no renegotiation, no reduction in service, no change order. Zero dollars of that $17 billion survive contact with the record. This is the pattern that matters more than any single number. A $17 billion fabrication is not a rounding error. It is larger than the annual budget of several federal agencies. If one line item can be invented at that scale, then the claim that "the totals are roughly right" loses all credibility. This is not random error. Error has a distribution; this pattern has a direction. Every distortion points upward. In my line of work, when a claimed figure is wrong, it is wrong in the direction of the speaker's incentive. I call the structure of this particular failure the triple distortion, and once you see it, you see it everywhere. First, target displacement. This is taking credit for reductions that would have occurred anyway. The 108 leases already tapering before January 20 are textbook displacement. The crypto equivalent is a lending protocol that counts deposits made before its launch as TVL growth — a vanity metric built on assets it neither sourced nor earned. Second, the information black box. Publish numbers without the reference points needed to verify them. The 96% of grant savings with no reproducible computation. The quarter of contract records with no identifiers. This is not sloppy record-keeping. It is a dashboard engineered for display, not for reconciliation. A ledger should be traversable. This one dead-ends everywhere. Third, narrative attribution. Assigning outcomes to actions that never took place. The Defense Health Agency's $17 billion is pure attribution — a line item assigned to contracts that were never touched. In crypto terms, it is a team citing revenue from a token that was never minted. It is not an error. It is the construction of a fictional transaction alongside the absence of a real one. Code does not lie, but incentives do. The Receipt Wall was not built for audit. It was built to produce a number that a presidential administration could campaign on. One incentive — show savings, show them fast, show them enormous — and it produces exactly the pattern above. I have seen this in a dozen protocol reviews. When the KPI is the product, the KPI gets fortified, and the truth becomes the casualty. The design made the failure more likely, not less. DOGE had no appropriations relationship, no civil-service continuity, no standing audit exposure. It was a flash governance vehicle — the DAO equivalent of a multisig with one signer and a press team. Any assertion not enforced by code is not an assertion; it is a proposal. And this year, as I audit the smart-contract interfaces of AI-agent platforms, I keep seeing the same structure: probabilistic systems publishing deterministic-sounding claims. DOGE is simply the highest-budget version of that error. The GAO's credibility, by contrast, is the product of being institutionally boring. It publishes methodology alongside conclusions. It responds to statutory mandates. It testifies before Congress. DOGE declined to respond to the GAO's requests for information and interviews. When the auditor asks for logs and the auditee goes silent, the audit gets shorter and harsher. In the security trade, we call that a scope limitation and write it into the report politely. It means the client refused to show the logs. The finding writes itself. The exploit was in the trust, not the contract. Let me be precise about what the GAO actually proved. It did not conclude that zero dollars were saved. That distinction matters. Auditors do not litigate; they express levels of assurance. What the GAO proved is that the claimed $110.3 billion cannot be substantiated at scale. Even under the most favorable reading — full credit for the verified contract terminations, minimal credit for the small fraction of grants that carry enough documentation to compute, full credit for the $31.8 million in leases — the verifiable total lands near $28 billion. A quarter of the claim. The Receipt Wall just received its first independent pricing event, and it traded at seventy-five percent off face value. The word "savings" also deserves more scrutiny than it typically gets. Terminating a contract does not automatically produce savings. Termination can trigger early-termination fees, litigation, or increased costs in downstream programs. The GAO verified, at best, that certain contracts were actually terminated. It did not verify the cash flows. The true fiscal effect is almost certainly smaller than even the verified terminations imply. Step back now, because magnitude matters. $110.3 billion sounds enormous. It is not. The federal budget runs between $6.5 trillion and $7 trillion per year. The national debt is $36 trillion. Even at one hundred percent face value, DOGE's claim would be less than two percent of annual outlays. At the GAO's implied valuation, it is about half a percent — close to noise. This guts the fiscal narrative. For months, the efficiency revolution has been used as a proxy for deficit reduction. The GAO confirms that the arithmetic never justified that proxy. If anyone was pricing US fiscal tightening based on the Receipt Wall, they were pricing a fabricated ledger. That is not a forecast about Treasury yields. It is a warning that a key variable was never in the model. The second-order damage is institutional. A government that cannot substantiate its signature savings claim loses data credibility. If the Receipt Wall was inflated by roughly 75%, what else in the official statistical apparatus is inflated? That ambiguity lifts the risk premium on federal data products — budget projections, procurement reports, employment figures — even if the move is marginal. Entropy always wins if you stop watching, and the federal ledger is too large for anyone to watch all of it at once. Read this now as a market event, because the report is, functionally, a short squeeze on the DOGE-cuts trade. Three sectors spent 2025 pricing aggressive federal downsizing: defense and government-services contractors, Washington DC commercial real estate, and health-IT vendors. The GAO findings say most of those cuts were narrative. Defense Health Agency contracts untouched. Lease savings seventy-two percent fictional. More than half of the marked contract terminations unreal. The logic held until the liquidity dried up. The GAO just confirmed the liquidity was never there. Repricing, though, is not an all-clear. DOGE's failure to execute its claimed cuts during its shortened life does not mean a successor program will fail to execute real ones. The next version will have better data hygiene and a much higher evidence bar — the GAO just set it. If a future savings dashboard survives audit, the impact on contractors and Washington office REITs becomes real, not rhetorical. The trade to price is the intervention that actually executes, not the one that was imagined. I do not write takedowns for the pleasure of watching wreckage. The Receipt Wall has a case, and the case deserves an honest accounting. First: transparency, even flawed transparency, beats the alternative. Most federal agencies publish spending data that is delayed, incomplete, and never reconciled in public. DOGE published claims quickly, itemized them, and made them falsifiable. That choice created the conditions for the GAO report. An unaudited dashboard is still more honest than no dashboard, because it invites the test that will expose it. Second, the underlying target is real. The federal government does waste tens of billions of dollars every year on redundant contracts, underutilized leases, and duplicative IT. A claim that lands at a quarter of face value still implies twenty-five to thirty billion dollars in identifiable, partially verified cuts. That is not nothing. The failure was in the accounting of the mission, not necessarily in the mission itself. Third — and this is the uncomfortable one — the GAO report is proof that the system's self-correction worked. DOGE refused to respond. The GAO audited anyway. A private billionaire's performance theater collided with an institution that has no token to dump and no opinion to sell. That is the longest-running audit-after-launch the federal government has ever conducted on a Musk entity, and it worked precisely because the auditor was institutionally boring. There is a second contrarian point, this one for macro readers. The collapse of the savings narrative removes an austerity tail-risk that was never real. If the market believed the federal government could find $110 billion in savings without touching entitlements, it might have modeled a cleaner deficit path. The GAO report unwinds that assumption. Treasuries do not rally on fake savings, but they also do not promise fiscal restraint on the strength of a Receipt Wall. The bulls were right that publishing the claims beats hiding them. They were wrong that publishing is the same as accountability. And the deepest bull case: the hostile audit has given the efficiency program something it never had — a benchmark. The GAO's refusal to rubber-stamp the number means future programs will be measured against a document, not a tagline. That is an upgrade. DOGE is gone. The Receipt Wall remains, like a smart contract with its owner revoked and its funds stranded. The GAO report is the transaction that confirmed the status: reverted. The path forward is not to abandon efficiency demands. It is to demand receipts that can be cashed. In Washington or in DeFi, the evidence bar just moved: a savings claim survives only if it can be reconstructed from public data alone. Otherwise it is narrative, not a number. Watch three things from here. First, whether Congress turns the GAO's findings into a hearing or a statute. Second, whether the Treasury's actual outlays corroborate even the verified sliver of the claim. Third, whether the next efficiency initiative publishes data that survives reconstruction. Each is a fork. None of them depends on the Receipt Wall. Logic is cold, but math is absolute. The only number that matters is the one that holds when the auditor stops reading the headlines and starts reading the logs.