DOGE vs USAID, Crypto Framework, Google's $75B AI Spend, US Sovereign Wealth Fund, GLP-1s
Summary
DOGE’s central bet is that zero-based budgeting can expose enough federal waste to narrow a $2T annual deficit and eventually relieve pressure on rates. Antonio Gracias cast the Twitter turnaround—roughly 80% fewer employees, payments frozen, unused vendors flushed out—as the rehearsal for government, while David Friedberg called DOGE “what we needed.” Gracias estimated fraud might exceed 10% of the $6.5T budget, but the episode offered that as a working hypothesis, not a verified saving.
USAID became the test case after its leadership allegedly resisted President Trump’s 90-day foreign-aid pause, prompting DOGE to follow the organizations violating executive orders most. The hosts cited a $45B Biden-era budget and grants including $2.5M for Vietnamese EV charging and $70K for an Irish DEI musical, while repeatedly acknowledging that breaking-news claims had not all been verified. The durable argument was Friedberg’s question: “What is the essential role of government?”
The most consequential DOGE allegation is that federal payment plumbing lacks the controller and reconciliation functions standard in any serious company. Gracias described money moving from congressional authorization through OMB and an agency request to Treasury without dependable verification that contracted goods or services arrived. His warning was stark: bureaucracy risked becoming “a Latin American-style kleptocracy,” though he conceded that nobody appeared to have full command of the system yet.
The USAID fight is becoming a legitimacy crisis for media and NGOs because undisclosed government funding can resemble political influence even where subscriptions or grants are legitimate. David Sacks compared the alleged loop to payola; Jason Calacanis cited federal Politico spending rising from about $1.3M annually during Trump’s first term to $8M under Biden, while cautioning that the figures still required verification. Sacks’s broader framing was “AstroTurf”: supposedly grassroots politics financed from Washington downward.
Crypto’s actionable policy sequence is stablecoin legislation first, followed quickly by a market-structure bill that could pass within several months. Sacks said all four relevant House and Senate committee chairs were aligned, with an updated FIT21-style framework defining currencies, securities, commodities and collectibles; the prior House bill received 71 Democratic votes. His consumer-protection thesis was to bring projects onshore, where disclosure and supervision can distinguish compliant founders from offshore frauds.
A US sovereign wealth fund divided the panel between industrial policy and disciplined asset monetization. Chamath and Gracias favored expert governance that could reinvest newly created assets—potentially a 50% TikTok stake valued by Chamath around $150B—and energy revenue into strategic American industries. Friedberg opposed borrowing at roughly 5% to become a venture investor, preferring a professional vehicle that sells TikTok, seized Bitcoin, land and similar assets, then returns cash to Treasury.
Google’s $75B 2025 capex plan looked more bullish than reckless once framed through utilization and return on invested capital. Revenue reached $96.5B for the quarter, net profit roughly $26.5B, yet the stock fell 7% as capex came in 29% above expectations. Friedberg calculated that a 20% return plus six-year amortization requires roughly $27B in incremental annual operating profit—substantial, but plausible for a company the panel considers unusually skilled at infrastructure deployment.
GLP-1s may become a broad health platform, but Friedberg wants to separate drug effects from weight loss, protein intake and resistance training before taking one himself. He cited VA cohorts totaling roughly two million diabetes patients, with GLP-1 users showing about 30% less cardiac arrest and lower rates of shock, hepatic and respiratory failure, and schizophrenia, against modest increases in gastrointestinal and related complications. The unresolved tradeoff is muscle and bone-density loss: “I don’t want to kind of confound the two factors.”
Deep dive
1. DOGE is applying the Twitter turnaround playbook to government
Gracias argued that “the DOGE story maybe starts with the Twitter takeover.” The acquired company carried roughly $12.5B of debt and $1.5B of annual interest, yet after about 80% of employees departed it could service that debt; a bank deal priced near 97 cents and subsequently traded above 98. He also said external brand-safety checks now rate the company at 99%.
His specimen of organizational decay was physical: conference-room pens had dried because nobody used them, while impeccable flowers and freshly prepared food for thousands were discarded three times daily. Calacanis added that roughly 20 San Francisco lunches effectively cost about $800 each after demand collapsed but supply did not.
The turnaround technique was brutally diagnostic: “Get the checkbook and just turn payments off, and then see what happens.” When credit cards were disabled, callers surfaced subscriptions and software nobody had activated; Gracias’s heuristic was that the loudest complainants could be “the worst grifters in the whole game.”
Government raises the stakes: Gracias put annual spending at $6.5T against $4.5T of receipts, with a $2T deficit and $1T of interest. Musk’s public 10% fraud estimate “might be low,” he said, while placing possible waste around $650B to $1T and warning that bureaucracy could harden into kleptocracy.
2. Zero-based budgeting forces the real debate over government’s role
Chamath rejected the idea that DOGE is unprecedented. The bipartisan Truman Committee began in 1941 with a $20,000–$30,000 budget, spent under $1M across roughly six years, and was estimated to save $1B–$15B then—up to about a quarter-trillion dollars in 2023 terms.
His second precedent was the Clinton-era National Partnership for Reinventing Government. Both prior efforts were driven by Democrats, making today’s categorical partisan opposition historically odd; Chamath’s conclusion was that established committees and inspectors general no longer provide timely enough accountability.
Friedberg supplied the operating model: take every expense “down to the studs,” rebuild from first principles around this year’s objectives, and never treat last year’s budget as an entitlement. That naturally surfaces the democratic question beneath every line item: “What is the essential role of government?”
Chamath described DOGE as “read-only auditors of the truth”: access lets them extract and publish information, not alter it. His claimed advantage over a congressional inquiry was speed—public data immediately, rather than selected findings after six to nine months.
3. Federal payment plumbing lacks a normal controller function
Gracias admitted, “I’m not sure I have full command of it. I’m not sure anyone does quite yet.” His provisional map ran from congressional authorization to OMB allocation, then an agency request and Treasury payment—with no dependable controller verifying contracts, purchase orders, delivery, payables and reconciliation.
He traced part of the constraint to 1973, after leaving the gold standard in 1971, when Congress removed presidential impoundment power because Nixon had used it to halt disfavored spending. The modern consequence, in his telling, is that an executive can be responsible for spending while having limited power to stop it.
This missing reconciliation supposedly explains why totals are revised and agencies cannot pass audits: “You can’t audit something you haven’t reconciled.” Gracias said the only audit he had seen was from the Social Security Administration, and even that report was “riddled” with material weaknesses.
His investment experience reinforced the concern. After encountering persistent Medicare and Medicaid fraud in services businesses, Valor adopted a rule against investments where those government payers exceed roughly one-third of the business—a private-market response to controls he considered structurally unreliable.
4. USAID attracted DOGE by allegedly resisting the foreign-aid pause
Calacanis framed USAID as a 1961 agency with at least 10,000 employees and programs in 130 countries as of 2023. Its budget rose from $26B to approximately $45B under Biden—about $150 per American—versus roughly $15B–$20B during Trump’s first term.
Trump paused foreign aid for 90 days on January 21. The White House subsequently alleged that USAID leadership was circumventing the order; Musk’s explanation was that DOGE checked which organizations were violating executive orders most, “turned out to be USAID,” and made it the priority.
According to the account presented, USAID security personnel then tried to block DOGE from its building or systems before access was obtained. The panel treated resistance as a warning sign, borrowing the Twitter lesson that whoever fights payment scrutiny hardest may have the most to protect.
Calacanis cited $2.5M for EV charging in Vietnam, $2M for sex changes and LGBTQ activism in Guatemala, $1.5M for a Serbian LGBTQ group, and $70,000 for an Irish DEI musical. He explicitly cautioned that much of the story was breaking and “hasn’t been verified yet.”
5. The media-money loop threatens legitimacy more than any one grant
Sacks’s historical analogy was payola: record labels once paid DJs to play songs, leading to legal changes requiring disclosure. He suggested government money reaching news organizations might be acceptable if disclosed, but questioned whether intermediaries responsible for informing the public had remained independent.
Calacanis said all-agency Politico spending averaged around $1.3M annually during Trump’s first term, then climbed to $8M under Biden; the circulating $34M figure covered years back to 2008. Against Politico’s roughly $200M revenue when acquired in 2021, $8M would have represented about 4%.
Other figures he flagged for investigation included $2.7M to the BBC in 2023—described as 8% of its annual income—and $120M since 2011 to Thompson Reuters’s consulting arm, roughly half during Biden’s presidency. His refrain was not guilt but verification: some publication spending is normal, while the acceleration looked suspicious.
Sacks widened the charge beyond media, saying it appeared that USAID financed political opposition in Hungary and Poland and that “every left-wing organization in the world” seemed to be receiving money from the slush fund. His political description was “AstroTurf”: top-down Washington money funding groups publicly presented as grassroots.
6. Foreign aid still has a constituency, but priorities and intervention split the panel
Calacanis defended the underlying aspiration: the West should act jointly to reduce suffering, and America can lead efforts against AIDS, poverty, disaster and severe rights violations. His objection was unilateral manipulation and pet-project funding, not humanitarian aid itself.
Drawing on his early Amnesty International work, Chamath contrasted imprisonment, torture, rape and murder with the organization’s later emphasis on trans rights. His proposed triage put systematic violence first and someone feeling misgendered “way down the priority list,” while acknowledging a principled role for human-rights assistance.
Calacanis said Sacks had previously explained the political-history puzzle by arguing that neoconservatives captured foreign policy in both parties and replaced older anti-intervention instincts with an activist posture.
Chamath added that some supposedly urgent social problems may prove embellished by funding cycles, making past cancellation campaigns look foolish in hindsight.
7. DOGE is remaking coalition politics even as courts constrain access
Chamath corrected his own earlier forecast that unreformed Republicans would lose for 10–15 years. Trump, he now argued, turned a conventionally pro-capital party populist and pro-labor, while Biden-era asset gains, deficits, immigration and wage suppression materially favored capital despite Democratic rhetoric.
Friedberg cited Rahm Emanuel’s warning that Democrats had abandoned “kitchen-table issues.” Inflation punishes wage earners, savers and retirees while owners of productive assets are protected; in Friedberg’s formulation, activists who believed they were fighting wealth were actually “feeding the rich.”
A federal temporary restraining order initially barred DOGE from accessing Treasury payment data. Reading it live, the panel revised its reaction: access appeared narrowed to two people, Tom Krause and Marko Elez, which sounded “not unreasonable” as a control, though two reviewers could materially slow a 24/7 audit.
Trump’s tax framework arrived during the discussion: no tax on tips, senior Social Security or overtime; renewed middle-class cuts; an adjusted SALT cap; and elimination of special breaks for billionaire sports-team owners and carried interest. Calacanis said Trump was “going for the jugular” because few politicians could publicly defend those preferences, while Chamath supported the plan.
8. Crypto’s first wins are stablecoins and market structure
Sacks said Trump’s week-one order established the mandate to support “responsible use and growth of digital assets and blockchains across every sector of the economy.” The legislative novelty is alignment among all four relevant committee chairs across House Financial Services, Senate Banking and both agriculture committees, which oversee the CFTC.
The sequence is stablecoins first, then market structure “very quickly.” Senator Hagerty had released a stablecoin bill, while Sacks expected a revised FIT21-style framework to define currencies, securities, commodities and collectibles—and how a project might begin as a security before decentralizing into a commodity.
The previous House market-structure bill received 71 Democratic votes but died in the Senate under Banking chair Sherrod Brown. With Tim Scott now chair, Sacks thought passage within roughly six months was plausible; outside reconciliation, Republicans would likely need seven Democrats to reach 60 votes.
His consumer-protection answer began with geography: bring innovation onshore so regulators and markets can identify compliant projects. FTX being based in the Bahamas was “a little bit of a tell”; clear US disclosure rules would make offshore holdouts “stick out like a sore thumb.”
9. Enforcement without rules drove legitimate crypto founders offshore
Sacks called Gary Gensler’s claim that the SEC welcomed crypto companies “very disingenuous.” Founders said they entered meetings seeking guidance, received none, watched enforcement staff record their statements, and then received Wells notices—“honeypotting founders,” in his description.
The replacement has two tracks: Commissioner Hester Peirce’s SEC task force can start clarifying treatment administratively, while Congress creates durable statutory definitions. The bargain is reciprocal: founders must comply, but government must disclose rules before prosecuting violations.
Reconciliation remained an uncertain possibility. Sacks noted that the Byrd Rule requires predominantly budgetary impact, though the Biden administration had stretched that boundary for the Inflation Reduction Act; he also emphasized that the proposal could attract bipartisan support if seven Senate Democrats join.
10. Washington’s AI reset treats China as the binding constraint
Trump rescinded Biden’s roughly 100-page AI executive order, which Sacks called a “monstrosity” of burdensome regulation. DeepSeek strengthened his case: China had either caught up or moved “very, very close,” invalidating the premise that America could regulate its companies as though no serious competitor existed.
The replacement is a new AI action plan under development. Sacks’s causal claim was direct: restrictions that slow US companies accrue to China’s benefit, so the policy goal is competitiveness rather than rules designed “in a vacuum.”
His late-night DOGE anecdote made the same cultural point: young engineers in suits and ties were working so late that facilities staff needed new access procedures because nobody usually requested them. “You’re coming to the office and doing work—we don’t have a protocol for that.”
11. A sovereign wealth fund needs assets and governance, not borrowed ambition
Trump’s order gave Treasury and Commerce 90 days to propose funding, strategy, structure and governance. Unlike resource-funded Norway, Saudi Arabia or the UAE, the initial American asset could be a negotiated 50% TikTok stake; Chamath estimated it might be worth roughly $150B “minted effectively overnight.”
Chamath proposed five unpaid, rotating billionaire stewards spanning markets—examples included David Tepper, Stan Druckenmiller, Ken Griffin, John Doerr, Mike Moritz and Bill Gross—who would deploy TikTok proceeds, federal-land sales and energy revenue into American companies. Governance, not whether the fund exists, was his decisive variable.
Gracias favored the fund as “a stealthy way to create industrial policy.” Instead of bureaucrats allocating a claimed $200B through programs such as the CHIPS Act, expert investors could back strategic industries economically, borrowing China’s long-horizon capital model without abandoning returns.
Friedberg’s objection was that the government is bad at capitalism. At a roughly 5% cost of capital, he opposed borrowing to invest; he preferred professional monetization of TikTok, land, seized Bitcoin and other assets, with proceeds ultimately returned to Treasury. Separately, he suggested considering smarter investment of Social Security assets.
Friedberg also claimed Social Security was functionally about eight years from bankruptcy under its current structure. Chamath called its $6T share of the $36T debt a “fake Treasury” sitting on a paper ledger and argued that productive investment could be better, while noting that the Social Security Administration was the only government entity they had found with an audit.
The panel disagreed in real time over seized Bitcoin: Calacanis said 144,000 Bitcoin were sitting somewhere in the Department of Justice, Chamath said they had been sold, and Calacanis said he thought they remained available for a proposed Bitcoin Strategic Reserve.
12. Google’s $75B capex looks like an ROIC bet, not blind AI spending
Google fell 7% despite quarterly revenue of $96.5B, up 12%; cloud revenue of $12B, up 30%; YouTube advertising of $10.5B, up 14%; and approximately $26.5B of net profit, up 28%. Full-year revenue was about $350B with $100B in net profit, while YouTube and Cloud ended 2024 with a combined run rate of $110B.
The concern was $75B of 2025 capex—42% above 2024 and 29% beyond analyst expectations. Chamath called Google’s underlying models “probably the best” across broad capabilities, while conceding OpenAI’s deep-research product was faster and marginally better.
Chamath’s diagnosis was disclosure: distinguish high-return AI improving ad targeting from speculative pre- and post-training, and investors could price the spending rationally.
Gracias said “models are basically commodities” in an increasingly competitive melee; data-center ROIC will separate winners. He argued xAI’s dense 100,000-GPU cluster was built unusually quickly and cheaply, while Google’s own chips, infrastructure discipline and advertising monopoly also position it to win.
Friedberg reconstructed Google’s depreciation progression from two-to-three-year servers, through four and five years, to six years after a 2021 change enabled by AI-based internal infrastructure management. At 20% ROIC, $75B needs about $15B of annual profit plus roughly $12B of amortization—around $27B total, a credible hurdle relative to Google’s base.
13. AI displacement is real, but productivity is the bullish counterweight
Gracias refused to dismiss transition costs: research on Pittsburgh steel losses in the 1970s put the economic cost around $1M per displaced worker because retraining often failed. “You can’t retrain” is the hard case that optimistic automation stories must absorb.
His offset was arithmetic: GDP growth is people working multiplied by productivity. If AI drives productivity to 5%, 6%, 7% or 8%, the resulting growth could fund new services, startups and jobs. With application-layer barriers low, he suggested ventures such as Launch could expand as displaced workers start companies.
Calacanis then noted that some companies are reaching $1M in revenue with five employees rather than the 25 that might previously have been required.
Chamath invoked Buffett’s account of industrial farming eliminating roughly 90% of agricultural jobs while the surrounding economy created unforeseen occupations. Closet organizers, life coaches, podcasters, influencers and venture capitalists are all “excess economy” jobs that would have made no sense during the agrarian transition.
His darker concern was that Americans have had too few recent repetitions at genuine creativity. A BYD car that parks itself with a little swipe seemed shocking partly because Chamath wondered why the capability did not exist in America; recovery requires unshackling builders from distractions and letting “these creative people cook.”
14. VA data made GLP-1s look broader than weight-loss drugs
Friedberg described a VA analysis comparing roughly 1.2M diabetic patients taking no medication, 215,000 taking GLP-1 receptor agonists, and 600,000 taking other diabetes drugs. He presented the cohort segmentation as a way to isolate differences associated with GLP-1 use across many diagnoses.
Reported downsides were comparatively narrow: roughly 8%–10% increases in nausea and vomiting, plus musculoskeletal complications, gastric reflux and indigestion-related sleep disturbance. The benefit side included lower shock, hepatic and respiratory failure, cardiac arrest and schizophrenia, with cardiac-arrest probability down about 30%.
Asked whether weight loss alone explained the results, Friedberg said emerging data showed effects in non-obese users too. His proposed mechanism was receptor-driven gene-expression cascades that suppress inflammatory markers and activate what he called “C2 genes” involved in cellular repair, though longer-term indication studies were still coming.
He would not put his children on a GLP-1 but would consider one for himself or his wife. Muscle and bone-density loss remain the counterweight, so his sequence is more protein and rigorous weightlifting first, then deciding whether to add the drug: “I don’t want to kind of confound the two factors.”
15. Health optimization’s bottleneck is a protocol people can follow
Chamath’s complaint was cognitive load: Gary Brecka, Andrew Huberman and Bryan Johnson each offer overlapping but meaningfully different protocols, leaving a busy person to reconcile salts, methylated vitamins, amino acids and supplements. “What’s the real bang for your buck?” was the practical question.
Multiple doctors and comparison layers had cost him more while making his care worse. What he wanted was a simple instruction set—whether to take metformin, vitamin D and omega-3s, what meal plan to follow, and which preventable actions matter most.
The concern was personal rather than cosmetic: Chamath’s father and best friend died after poor health, so he becomes obsessive about controllable risks.
Sacks suggested Friedberg document his weight, muscle-building efforts and any later GLP-1 use as an everyman’s longitudinal journey, giving confused consumers a weekly account they could actually follow.