Hot I.P.O Summer + What Is A.I. Doing to Math? + HatGPT

What It Covers

SpaceX, Anthropic, and OpenAI are simultaneously pursuing what could be the three largest IPOs in history, potentially reshaping San Francisco's wealth distribution and AI governance. Journalist Kevin Hartnett joins to examine how AI is transforming mathematics research, from solving century-old Erdős problems to triggering a backlash from 800 mathematicians signing the Leiden Declaration.

Key Questions Answered

• •AI IPO Scale: SpaceX plans to sell shares at $135 each, targeting $75 billion raised and a valuation between $750 billion and $2 trillion — potentially the largest IPO in history. Anthropic, which had roughly $1 billion annualized revenue in January 2025 and recently reported $50 billion run rate, and OpenAI are filing S-1s simultaneously, creating an unprecedented concentration of public market debuts in a single year.
• •Philanthropic Capital Surge: Anthropic's eight cofounders pledged to donate at least 80% of their wealth to charity. The company also offered early employees a three-to-one stock match for charitable pledges. The resulting philanthropic capital flow could exceed the Gates Foundation's annual giving, directed primarily toward AI safety, global health, pandemic preparedness, and effective altruism cause areas including animal welfare.
• •AI Safety Risk Under Public Markets: Both OpenAI and Anthropic are structured as public benefit corporations, which reduces shareholder lawsuit exposure for safety-motivated decisions. However, going public introduces activist investors and index fund pressure that could accelerate model releases. The counterargument: releasing a genuinely dangerous model exposes companies to securities fraud liability from shareholders who relied on safety commitments.
• •AI Math Milestone — Unit Distance Conjecture: OpenAI's model disproved the unit distance conjecture, an Erdős problem mathematicians had actively studied for decades. Unlike prior Erdős solutions dismissed as sophisticated puzzles, this proof used novel, non-obvious methods and was assessed as publishable in the Annals of Mathematics, the field's top journal. This marks the first broadly accepted evidence that AI can produce top-tier original mathematical research.
• •Leiden Declaration — Mathematician Resistance: Over 800 mathematicians, including Terence Tao, signed the Leiden Declaration expressing concern that AI generates plausible but unverifiable proofs, distorts research incentives toward problems AI handles well, and erodes human authorship norms. The math arXiv preprint server now bans users for one year if they upload unedited, unreviewed AI-generated proofs detected through metadata analysis.
• •Three Camps Among Mathematicians: Mathematicians currently split into three positions on AI: outright rejection after encountering model errors, cautious optimism represented by Terence Tao's "jetpack for thinking" framing where AI accelerates hypothesis testing, and a minority view that AI will make human mathematicians obsolete within two years. The Tao middle-ground position — humans directing AI toward problems, AI executing — appears to hold the most support within the field currently.