The Frontier

SpaceX plans to sell shares at $135 in its upcoming IPO, raising $75 billion. This would value the conglomerate at $1.75 to $2 trillion, making it the largest IPO in history.

Kevin Roose describes SpaceX as a conglomerate containing two strong businesses, reusable rockets and Starlink internet, and two weaker ones, XAI and X, the social network. He notes XAI now rents compute to Anthropic.

Anthropic filed a confidential S1 to go public. Kevin Roose notes its annualized revenue run rate hit $1 billion in January 2025, representing unprecedented growth from its early days as a small safety-focused research group.

Casey Newton argues the impending IPOs of Anthropic and OpenAI will exacerbate inequality in San Francisco. He says mid-six-figure earners now feel precarious compared to potential millionaires minted by these companies.

Anthropic's co-founders pledged 80% of their wealth to charity, and the company matches employee equity pledges to philanthropy 3-to-1 for early hires. This will create a massive influx of philanthropic capital after the IPO.

Kevin Roose worries that going public will pressure AI companies like OpenAI and Anthropic to prioritize speed over safety, as public markets and activist investors demand growth.

Casey Newton counters that public markets could introduce democratic oversight, forcing companies to disclose financials and earnings, providing levers for accountability that don't currently exist.

In May 2025, an OpenAI model disproved the unit distance conjecture, a significant geometry problem on Paul Erdős's list. Kevin Hartnett says mathematicians viewed this as a top-tier research result publishable in top journals.

Kevin Hartnett says mathematicians hold three views on AI: dismissive, fearful it will make them obsolete, and optimistic it acts as a cognitive jetpack. He believes the optimistic Terry Tao camp is currently winning.

Over 800 mathematicians signed the Leiden Declaration, expressing deep concern that AI-generated proofs are hard to verify and could erode the human foundations of the discipline. The arXiv archive now bans users for a year if unedited AI text is detected.

President Trump signed an executive order asking tech companies to voluntarily give the government a 30-day review period for new AI models before release, a reduction from an earlier 90-day draft requirement.

A San Francisco startup, The Bot Company, is being sued for $12,383.50 after secretly using an Airbnb to train robots, which allegedly damaged the property and violated rental terms.