World's First Trillionaire, Anthropic Fable Banned, The New Oligarchs, Iran Peace Deal

Elon Musk became the world's first trillionaire following the SpaceX IPO, which priced at 135 dollars per share, raised 85 billion dollars including the green shoe, and was three times larger than the Saudi Aramco IPO in 2019. SpaceX stock closed up 19 percent on IPO day and was trading at 177 dollars at time of recording, briefly pushing the market cap above 2 trillion dollars and making SpaceX the fourth most valuable company, passing Amazon and Microsoft. Musk's trillionaire status reflects paper wealth from a higher public valuation of shares he already owned, not new cash, and he is subject to a one-year lockup. The IPO allocated 20 to 30 percent of shares to retail investors, but current SEC rules had barred the bottom 95 to 96 percent of Americans from buying SpaceX stock at valuations of 10 billion, 100 billion, and 500 billion dollars. SpaceX then exercised its option to acquire Cursor, valuing it at 60 billion dollars, roughly 15 times its approximately 4 billion dollars in annual revenue, with Chamath arguing the stock-for-stock structure gave Musk an effective 50 percent discount as SpaceX's valuation grows post-IPO.

Anthropic's model internally called Mythos was held 30 days for cybersecurity review before being released publicly as Fable 5 on June 9th, then shut down days later. Commerce Secretary Howard Lutnick directed Anthropic to restrict Fable 5 to US citizens only, and Anthropic could not or would not implement sufficiently granular geographic restrictions and pulled the model globally. David Sacks said Dario Amodei came to Washington in April and described Mythos as a cyber weapon, raising national security concerns, and Anthropic then expanded the Mythos trusted partners program to approximately 50 or more companies without consulting the White House. The Washington Post reported Anthropic shared Mythos with a company the White House believed had connections to China, later identified by Wired as SK Telecom, which holds roughly 70 percent of South Korean telecom market share and carries longstanding allegations of ties to China. The White House had previously ordered Anthropic to revoke SK Telecom's access, and Sacks said Anthropic did not initially disclose to the White House that it had given Mythos to SK Telecom in the first place.

Private companies testing Fable, not the White House, first discovered the guardrails had been broken via a jailbreak. Amazon CEO Andy Jassy personally escalated the jailbreak to the White House after a communication breakdown with Anthropic, and the Treasury Secretary personally called Dario requesting Fable be taken down. Sacks said Dario refused until the jailbreak could be fixed and argued pedantically that it was not serious, then published a blog post attempting to differentiate minor from major jailbreaks, which Sacks described as off-brand for an AI safety company. Dario did not take the model down until after the White House sent a formal export control letter. Sacks acknowledged he has no firsthand knowledge of which specific company received Mythos improperly and is relying on Washington Post reporting, and said the action was not politically motivated and that he was not involved in the decision.

Chamath argued the incident creates an opportunity for hyperscalers to position themselves as government-approved gatekeepers for AI models, potentially compelling governments worldwide to require identity verification to access AI. He said neoscalers cannot replicate the KYC and VPC infrastructure Amazon, Microsoft, and Google have built over decades, meaning frontier lab mismanagement could produce an oligopoly controlling what he called the most economically leveling technology ever seen. Multiple speakers described a pattern in which Anthropic convinced Biden administration officials that AI was extremely dangerous and needed government control alongside a small number of hand-selected companies, then saw those officials go to work at Anthropic after that administration ended. A Claude-generated psychological analysis of Dario based on his own essays described his pattern as epistemic exceptionalism, finding that his framework outputs himself as the trustworthy key holder regardless of inputs.

Friedberg argued the AI stack will likely disaggregate similarly to computing in the 1980s, noting that an entire AI model can fit on a USB drive and that the value of any given model is time-bound because better models will replace it. He drew a parallel to IBM's control of the mainframe computing stack and argued government intervention forcing IBM to disaggregate its software layer started the independent software vendor era, predicting the same pattern will repeat with AI.

A formal signing of an Iran deal is scheduled for Friday June 19th in Geneva, mediated by Pakistan, following a conflict that began February 28th and lasted approximately 110 days. Iran commits to not developing nuclear weapons, agrees to destroy its enriched uranium stockpile under IAEA supervision, and freezes its nuclear program for 60 days, in exchange for all sanctions lifted, access to frozen assets, 300 billion dollars in reconstruction funding sourced entirely outside the US, and a commitment that US forces will withdraw from the region after a final deal. Three issues remain unresolved: Israel's sign-on, the question of Iran's nuclear enrichment going forward, and the status of Iran's ballistic missile program. Removing enriched uranium is estimated to set back any restart of a nuclear weapons program by years to a decade or fifteen years.