World's First Trillionaire, Anthropic Fable Banned, The New Oligarchs, Iran Peace Deal
In a Nutshell
The hosts argue that a new American politburo of progressive politicians and hyperscalers seeks to seize control of capital, production, and AI access under the guise of fairness and safety, while Elon Musk's SpaceX IPO and trillion-dollar valuation prove that free markets reward makers who build value-creating machines rather than takers who redistribute them. They condemn Anthropic's covert expansion of its Mythos cyber model to questionable partners and refusal to fix a jailbreak, which handed the government an excuse to impose export controls and positioned hyperscalers as gatekeepers requiring KYC for frontier models. On foreign policy, they call Trump's Iran peace deal—reopening the Strait of Hormuz, removing enriched uranium, and lifting sanctions without US reconstruction costs—a major achievement that avoids endless war and regime-change fantasies pushed by neocons.
These notes were generated by AI and may contain inaccuracies.
The All-In Podcast opens with the original quartet present. David Sacks had a busy week with long posts on X. Chamath mentions having one long post teed up for the podcast. The hosts discuss Freeberg's increased posting activity and how his tone has become more "based" as Opendoor's valuation has increased.
Freeberg describes what he calls the great American politburo being formed in the United States. The new oligarchs are arranging power structures, determining workforce allocation, and making decisions about capital flow. He identifies Elizabeth Warren, Bernie Sanders, and Ro Khanna as key figures attempting to coalesce power and create a system of greater control over the economy.
These politicians want to seize the means of production, control education, control media, and react negatively whenever an individual or enterprise succeeds outside their span of control. Freeberg argues they contort discussions about inequality, fairness, and justice while actually representing the rising empire that seeks to take away individual freedom and liberty.
Chamath asks Freeberg to expand on who comprises this politburo. Freeberg explains that in an unfree society, the politburo consists of leaders who elect themselves to control economic flow, capital allocation, and what activities individuals are permitted to pursue.
Freeberg states that the United States' power derives from giving people individual liberty, which unlocks agency to progress, educate oneself, make money, and gain assets and freedom. Economic mobility is the key value proposition of the United States.
When government provides jobs or money, economic mobility is lost and people become indentured to the government. Freeberg emphasizes that people sign up for promises of free education, childcare, food, and paychecks without realizing they are giving up their liberties and progress in exchange.
Chamath shares his personal story of growing up on welfare in Canada. His family received approximately $17,000-$19,000 annually. His father struggled with alcoholism and chose not to take available jobs paying $8-$10 per hour. Chamath describes the threshold for learned helplessness as far lower than people think, and warns that providing $30,000-$80,000 annually in government support will reduce people to shells of their potential.
Chamath's father eventually got a low-level clerical government job near the end of Chamath's high school years, which changed him completely. Chamath keeps a positive sense of his father despite the challenges.
The conversation addresses legislation that passed in Illinois (Pritzker's law) which the hosts claim effectively states there is no longer private property in the United States. The government can now take a piece of property annually after income tax has been paid and the asset has been purchased.
Freeberg explains that in socialist, communist, monarchist, or totalitarian systems, a small number of people ultimately own and control everything. He argues this is the brink the United States is approaching.
Sacks discusses the psychological concept of learned helplessness. When people are given impossible problems to solve followed by easy problems, they become depressed and internalize failure. When given progressively harder problems, people become empowered and maintain agency.
Sacks connects this to alcoholism, noting that when people attribute problems to themselves and self-medicate, the cycle becomes pernicious. He emphasizes the importance of tapering people off unemployment and giving them chances to be successful.
SpaceX had a record-breaking IPO at $135 per share. Robinhood, Charles Schwab, and others participated in allocations. The IPO raised $85 billion, three times what Saudi Aramco raised in 2019. The stock closed at $161, giving a market cap above $2 trillion. At the time of recording, the stock was at $177.
SpaceX exercised its option to acquire Cursor, the coding agent. Cursor had originally used Claude (from Anthropic) as its backend. Anthropic allegedly had an internal skunkworks project developing a coding agent despite telling Cursor they would never release one publicly. This led Cursor to develop their own model using xAI hardware at xAI.com. Cursor is doing $4 billion in revenue and was acquired for 15 times revenue.
SpaceX is trading at approximately 60-70 times revenue. SpaceX briefly passed Amazon and Microsoft to become the fourth most valuable company before settling as the seventh largest company behind TSMC.
SpaceX's 2025 revenue is projected at $19 billion, compared to Amazon's $217 billion and Microsoft's $282 billion.
Chamath notes that the acquisition was structured so the S-1 doesn't go stale. The deal is effectively done, and the announcement timing relates to avoiding rewrites of disclosures and risks.
Chamath calculates that if SpaceX is valued at $1 trillion today and could reach $2 trillion, Elon effectively received a 50% discount on the Cursor acquisition in future dollars. Additionally, the revenue run rate would double, meaning Elon got Cursor for effectively $15 billion.
Chamath calls the business intellect demonstrated "off the charts" and predicts a consolidation phase where Tesla and SpaceX will play significant roles.
Sacks explains that Elon Musk becoming the world's first trillionaire does not mean he has a trillion dollars sitting in a bank account. His balance sheet contains the same assets as before the IPO—he is not selling shares due to a one-year lockup and is predicted to hold longer because SpaceX is his life's work.
Sacks distinguishes between stuff and the machines that make stuff. Wealth comes from creating machines (corporations) that produce stuff for humanity over time. The discounted present value of all future production is what creates wealth today.
Stuff itself depreciates and is a wasting asset. The machinery that creates ongoing value is what matters. Sacks notes that Karl Marx failed to understand that there is no sharp delineation between capital and labor.
A welder at SpaceX reportedly made a million dollars in SpaceX stock through equity participation. The tech industry is more inclusive in terms of ownership than other industries.
Capitalism in free markets allows anyone to enter the market and transition from labor to capital. People can buy shares, start businesses, work, save, invest, and compound capital until their labor is worth less than their capital.
Freeberg argues that denouncing Elon's success tells young people they should not have the capacity to achieve similar outcomes, removing their agency to transition from labor to capital through sweat equity.
SpaceX is described as a machine that makes satellites.
SpaceX operates as a machine that makes satellites, broadband connections from space, launch vehicles, software, and AI software. The company creates things people want that enhance lives. Market value and public valuation of these products determine SpaceX's stock value and Elon Musk's net worth. Musk does not possess one more dollar than before the valuation increase. This wealth must be earned continuously by making the best product in each class. Starlink faces competition from two or three upcoming rivals. Other companies are developing rockets. Musk competes in LLMs against Cursor and Quad Code. He has had to fight to win each category, and stopping the fight means losing equity as seen when stocks tank.
This represents a 25-year overnight success story with three or four distinct points where the company faced going out of business. The notion that Musk won the lottery is farcical. People who criticize wealth accumulation are often secretly jealous. No coherent argument exists for dismantling successful achievements. Market fluctuations can make net worth numbers change dramatically. A version of the market unrelated to Musk could judge his companies at half their current value.
Machines become obsolete quickly in capitalist societies through disruption. Companies that become stale or calcified without investment in upgrades face obsolescence. Government represents the one unobstructed monopoly that cannot be disrupted. Government competes for capital and markets, which is why certain actors are described as evil. They function as a Politburo attempting to remove capacity for value creation that improves medicine, education, livelihoods, and homes across generations. When government monopolizes these areas, improvement stops.
Retail investors want earlier participation in company success. Elon Musk and the SpaceX team allocated 20 to 30 percent of the IPO to retail investors. Robinhood users received an interface offering participation in the SpaceX IPO, with six or seven hundred thousand users potentially receiving allocations ranging from one share to fifty shares. This democratization needs to continue so people can vote with their dollars for preferred futures.
People watching SpaceX rockets launch for years would have placed bets when the company was worth 10 billion, 100 billion, or 500 billion, but were prohibited. Current rules allow the top 4 or 5 percent of the country to buy private company stocks while deeming the remaining 95 percent too unsophisticated. This antiquated system must change. People should be able to buy crypto, private company stock, or engage in other financial activities of their choosing.
Joseph Schumpeter explained that the intelligentsia in society produces words and ideas rather than making things or the machines that create things. These ideas are often incorrect. The intelligentsia becomes resentful that creators of value-generating machines receive the greatest financial rewards because those machines align with society's greatest value. The great lie claims society divides between rich and poor with the rich being unfairly wealthy and the poor being unfairly poor, requiring redistribution. The great truth identifies makers and takers as the actual divide.
Makers include artists, plumbers, electricians, woodworkers, computer scientists, and people who build and create value for others across all income levels and wealth brackets. Takers include analysts, espousers, armchair mechanics, critics, commentators, and politicians who observe others creating value. Parents should ask children what they made each day and whether someone else valued it. If someone else values the creation, the maker has achieved something valuable and gained customers or partners.
The US government pulled the plug on Anthropic's latest model shortly after the previous episode taping. Anthropic released a controlled version called Mythos held for 30 days of security testing, then released Fable as a guardrailed version. Commerce Secretary Howard Lutnick instructed Anthropic to restrict the model to US citizens. Anthropic could not or would not implement granular restrictions and shut down the model entirely. This occurred days after the June 9th launch.
Amazon CEO Andy Jassy, an early investor with large ownership, reported a security vulnerability in Fable to the administration. A jailbreak bypassed Anthropic's guardrails. Dario Amodei stated the jailbreak was narrow and not serious. Semaphore reported White House action partly over suspicion that a China-linked group accessed Mythos improperly. Wired identified the group as South Korea's SK Telecom, which holds 70 percent market share in South Korean telecom and faces longstanding allegations of China relationships. The White House previously ordered Anthropic to revoke SK Telecom's Mythos access.
Dario Amodei visited Washington in April claiming creation of a cyber weapon called Mythos, elevating official concern about advanced cyber capabilities. Anthropic ran a trusted partners pilot program that expanded to approximately 50 companies without White House consultation. The Washington Post reported Anthropic shared Mythos with a company national security officials believed had China connections, specifically SK Telecom.
Anthropic conducted its own foreign policy by expanding the Mythos preview beyond White House knowledge. Once discovered, officials determined Mythos capabilities required export controls. Fable launched as Mythos with guardrails. Private companies testing Fable, including Amazon as Anthropic's largest shareholder and cloud partner, concluded guardrails failed with a serious jailbreak. Amazon's security team documented this as a potential liability and national security threat.
The White House attempted contacting Dario Amodei with multiple conversations. Dario refused requests to take down Fable until the jailbreak could be fixed. The Treasury Secretary called Dario personally. Dario argued the jailbreak was not a serious issue and published a blog post differentiating good jailbreaks from bad jailbreaks, minor ones from major ones. This contradicted Anthropic's safety-focused positioning.
The administration sent an export control letter requiring Fable to be taken down until guardrails were fixed. This represented a reaction to a credible national security threat report rather than new policy establishing government approval for every model release. Anthropic is now negotiating with the government to resolve the situation. Trust has been broken, making extrication more difficult than avoidance would have been.
Frontier lab leaders demonstrate consistent patterns of evasiveness and immaturity that damage the entire AI movement. AI represents the grand leveler enabling unique value creation and economic mobility for everyone. Instead, doomerism, hype cycles, and naivety create mistrust. Silicon Valley shifted from misfits building important things without ostentation to building less important things with negativity and mistrust.
Frontier lab leaders are public enemy number one. This episode proves these individuals cannot be trusted. The situation creates opportunity for hyperscalers to position themselves as gatekeepers to governments worldwide. Arguments will claim frontier labs cannot be trusted, models are uncontrolled, hyperscalers should provision models globally with KYC requirements, customer identification, and audit trails for prompts.
This creates an oligopoly placing the most powerful economic leveling instrument in the hands of a handful of hyperscalers. The counterfactual argument for open, distributed environments becomes difficult against compelling infrastructure arguments. Asking neoscalers to build equivalent KYC and VPC infrastructure to Amazon and Microsoft is impossible. Data centers and neoscaler markets could be removed from competition. Instead of diverse, robust, open ecosystems, debate now centers on gatekeeping, duopoly versus oligopoly.
Users may need to provide driver's licenses, tax IDs, or Social Security numbers to Amazon Web Services or Google to access models. Fable represents a precursor to knowledge that could be misused, similar to fertilizer purchase restrictions requiring identification.
Silicon Valley appears to have optimized for its own interests, taking itself seriously while disregarding others. This approach is viewed as unserious, immature, and unhelpful to broader society.
Anthropic stands out as the only major AI company absent from President Trump's White House CEO gatherings. The company is backed by Reid Hoffman, who funded lawfare efforts against the White House. This has created significant friction between Anthropic and the current administration. Questions arise about whether the White House is treating Anthropic differently and whether political motivations drive these actions rather than pragmatic concerns.
No regulatory framework exists for AI currently. Antagonizing the government to establish such a framework, then feeling antagonized when regulation occurs, creates a contradictory dynamic. Anthropic appears unique in this behavior among AI companies. The core concern is that government picking winners and losers in AI harms everyone, as does restricting access to these tools through government gatekeeping.
An entire AI model can fit on a USB drive with widely available hardware to run it. Model value is time-bound since better models will emerge. Having a single stack owned by one company requiring ID verification and government approval creates a losing scenario for all parties. Natural market forces are pushing toward more open source solutions and fragmentation of the AI stack.
In the 1960s, IBM dominated mainframe computers with a complete vertical stack including chips, hardware, operating system, and software. Government intervention eventually forced IBM to disaggregate the software layer, which initiated the independent software vendor era. This disaggregation enabled the software industry's growth. The 1980s brought further disaggregation with IBM using Intel chips and Microsoft developing superior operating systems, leading to proliferation of independent vendors.
Both mainframe computing and desktop PC eras drove productivity booms specifically because of disaggregation. The same pattern is expected for AI companies. Current oligopoly of three or four businesses is unlikely to persist. Hardware will feature multiple chip vendors, multiple clouds will exist, multiple models will run across clouds, and models will run locally. Users will employ different software factories beyond Anthropic, including tools like Salesforce and others, running purpose-built or custom applications.
If another company like OpenAI or Gemini released a model excelling at hacking and cyber warfare, limiting initial access to 50 partners including Microsoft and SK Telecom would receive approval. Holding the model for 30 days of testing would be viewed positively. Adding guardrails after a partner jailbreak would be accepted. Monitoring user activity during the first 30 days of public beta would be considered standard practice. Restricting access to logged-in customers and dropping users attempting dangerous queries like nuclear weapons or fertilizer regulation would be viewed favorably. Compliance with government requests to restrict access outside the United States would also receive approval.
A backdrop of political tension exists between Anthropic and the current administration. Reid Hoffman's connection and Dario Amodei's refusal to participate are noted. Anthropic attracts employees who oppose Trump and the current administration. In contrast, OpenAI and Grok attract personnel largely supportive of Trump, including Greg Brockman and his wife who contributed $25 million to Trump. Anthropic executed the Mythos rollout conservatively, but communication issues persist with Dario's frequent appearances involving excessive fear-mongering.
Industry self-regulation is essential. Google, Microsoft, Amazon, and Elon Musk should agree on testing standards and self-certify models before government involvement. This mirrors MPAA and video game industry practices. Self-certification would allow model releases without government intervention.
The Mythos restriction resulted from a jailbreak report and Dario declining to address it. The decision to expand the Mythos preview to a company considered a national security threat contributed to the outcome. This was not a dispute involving the Department of War or specific individuals. People sympathetic to Dario and Anthropic within the administration attempted to reset relations, but Dario's response undermined those efforts.
Pete Hegseth tweeted three months prior stating the Department of War had permanently removed Anthropic from the building, claiming each day validated that decision. The tweet timestamp occurred after the restriction letter was sent, reflecting vindication rather than involvement in the decision itself.
As an Anthropic shareholder, recommendations would include attending all Trump CEO gatherings and making political contributions comparable to Greg Brockman's $10-25 million donation. When a cabinet secretary reports a credible national security threat from a trusted partner, taking it seriously is essential. Taking the model down only after receiving the restriction letter demonstrated insufficient responsiveness.
A psychological analysis was conducted using Dario's essays Machines of Loving Grace and AI Policy on the Exponential. The analysis identified distrust of other labs for reckless racing, authoritarian states for weaponizing AI, markets for distributing gains, institutions for insufficient speed, and post-Mythos distrust of government transparency and fairness. Trustworthy actors appeared limited to those reasoning similarly under self-designed rules. This pattern suggests epistemic exceptionalism where one's reasoning is considered load-bearing and others' disagreement indicates their corruption rather than one's own error.
Anthropic's mantra positions AI as extremely dangerous with only themselves virtuous enough to control negatives. They view themselves as adjudicators of all relevant decisions. This superiority belief attracts talent but reflects a God complex where disagreement stems from others' error. The word misunderstanding in their Mythos communications implies that correct understanding would lead to agreement with their position.
Anthropic captured the Biden administration's AI policy apparatus. All Biden administration AI policy leaders, including NSC AI officials and the AI Safety Institute head that Anthropic helped establish, subsequently joined Anthropic. Anthropic organized government affairs early, with Dario engaging Washington to promote AI as extremely dangerous requiring government control through a merger between government and a small number of selected companies.
Mark Andreessen heard messaging that government would anoint AI winners with only a couple of companies receiving investment. Anthropic frames competition as a dangerous race condition where safety is compromised. They advocate centralization and control, viewing competition as nefarious. This approach seeks to create a cartel, appearing virtuous through safety language while serving self-interest. The fundamental philosophical divide concerns whether competition protects consumers, provides choice, improves competitors, prevents regulatory capture, and leads to decentralization, or whether centralization represents AI's greatest totalitarian threat.
After being removed from the Department of War following unsuccessful contract negotiations, Anthropic came to Washington in April declaring they had invented a cyber weapon. This alarmed officials and secured support for a government approval regime. Hubris led to missteps that resulted in being removed under their own proposed pre-approval system.
Newsweek and New York Times published 1961 articles predicting workforce demise from computing automation. Mainframe computers ultimately enabled workers to accomplish 100 times more work. Desktop computers similarly enabled economic mobility and productivity gains with new industries emerging. Federal government workforce training programs in the early 1980s spent hundreds of millions with no meaningful impact. Companies successfully helped employees transition to higher productivity using new tools rather than requiring massive retraining programs.
Technologists across the spectrum, from Elon Musk to Sam Altman to Dario Amodei, exhibit deep arrogance assuming this technological moment differs fundamentally from previous eras. Historical predictions of workforce elimination and job replacement proved incorrect. Technology continues as part of a long continuum of productivity improvements driven by human ingenuity that unlocks human potential. The job loss and end-of-world narratives create unnecessary frenzy. AI CEOs have engaged in unnecessary public fear-mongering about AI.
The Anthropic situation should not become precedential for policy-making. Emergency-driven policy is undesirable. Dario stated in an April Bloomberg interview with Emily Chang that withholding Mythos caused massive commercial damage. However, releasing Mythos without guardrails in April would have created massive legal exposure, as companies would have been hacked with foreseeable consequences. Meta recently lost a multimillion-dollar verdict in a body shaming lawsuit, and both Meta and Google face product liability lawsuits. Safety features were required for liability protection rather than representing a noble sacrifice for humanity.
The discussion addresses why hyperscalers would benefit from frontier labs being constrained. Hyperscalers have trillions of dollars in on-balance sheet and off-balance sheet exposure to AI. The best way for them to underwrite that investment is to kneecap frontier labs, play gatekeeper, charge a toll, and take a tax. This is described as business 101 and regulatory capture 101, where hyperscalers position themselves as adult supervision.
The frontier labs are described as having doom-trolled their way into a position where the public distrusts them and demands a policeman. This differs from traditional regulatory capture, which involves surreptitiously introducing bills in the dead of night. Here, the labs bumbled into a situation where they handed the keys to the hyperscalers, locking themselves up.
After 110 days of conflict that began on February 28th, President Trump announced for the 37th time that the war has ended. An initial agreement on June 15th will be codified with formal signing on Friday, June 19th in Geneva. The deal was created by Pakistan and extends the ceasefire for 60 days while also including a Lebanon ceasefire.
The US gets the Strait of Hormuz reopened. Iran commits to not developing nuclear weapons, agrees to destroy its stockpile of enriched uranium under IAEA supervision, and will freeze its nuclear program at current levels for 60 days. In exchange, Iran gets all sanctions lifted and $300 billion in reconstruction funds. The US is not paying for the reconstruction. Iran's access to frozen assets will be restored, and the US will remove its forces from the region after the final deal.
Unresolved issues include Israel's sign-on to the deal, Iran's nuclear enrichment going forward, and Iran's ballistic missile program.
David Sacks describes the deal as a tremendous, immense achievement for the president that was very difficult to negotiate given that Iranians are not easy to negotiate with.
The deal reopens the Strait of Hormuz so oil and vital materials flow for the global economy. There is a commitment from Iran not to pursue a nuclear program and to collect all nuclear materials. There is a ceasefire across all fronts. The reconstruction fund costs the US nothing, with Gulf states potentially involved in funding. The deal could lead to a rapprochement with Iran, which Sacks views as positive since the US does not need to be at war with that country forever.
Sacks criticizes neocons who want ground troops sent in for regime change in Iran. John Podoritz is cited as suggesting the volunteer army signed up for this, which Sacks rejects. Troops signed up to protect America, not to charge into Iran, described as a mountain fortress. Estimates suggest it would take over a million troops, and even then success is uncertain. The US sent half a million troops to Iraq, and Iran is three times bigger as a country.
Sacks states he is not sending his kids to fight that war and suggests if Podoritz and the neocons want it, they should send their own kids. If Reza Pahlavi wants to muster an army from Beverly Hills supporters, he should try, though Sacks doubts they will sign up given they are living well there.
Sacks argues that if ground troops are not being sent because it makes no sense, and if continued bombing is not pursued because all military targets above ground have already been flattened, then door number three—a peace deal—should be tried. The parties are at the MOU stage, and Sacks advocates giving peace a chance rather than signing up for another forever war like Iraq and Afghanistan.
The long-term issue of uranium enrichment to weapons-grade levels is addressed. The deal includes removal of Iran's enriched uranium stockpile from the country. If all enriched uranium comes out, even if Iran restarted the program, it would take years to a decade to 15 years plus to build enough enriched uranium for a nuclear weapon. This would provide a natural setback and defense for conditions that matter most.
The MOU stage means high-level principles are agreed, but exact program details still need definition. Iran gets sanctions relief and a more normalized relationship with the US in exchange.
Jason Calacanis asks whether going through the last couple of months was worth it if all Iranian enriched uranium is gone, missiles are depleted, and enrichment capacity is eliminated. He notes that as venture capitalists and capital allocators, they do not have complete information and must remain humble about what Trump sees.
Calacanis argues that Israel's trimming of the grass and hedging of the garden approach—isolating dictators and waiting them out—was the right strategy. Starting full-scale wars carries unknown escalation risks. He states that sending air troops periodically to clean up facilities would have been better, similar to how North Korea has been contained. The US did not need to start a full-scale war and will probably wind up where things started. Calacanis hopes the people of Iran eventually become free but emphasizes that democracy cannot be enforced in the Middle East—the people must lead their own revolution.
Keep All-In Podcast in your library
Save the videos and channels worth coming back to, and find them again in one place.
