Reading the Anthropic S-1, with Zvi Mowshowitz

Anthropic's S-1 filing reveals a founder-controlled structure with concentrated voting power, approximately $900 billion in forward compute commitments, and an expensive uncapped employee donation match program. The company's softer Q3 revenue reflects a shortage of GPUs rather than customers, raising questions about the sustainability of its massive compute investments and the broader AI industry's financial model. The filing also highlights risks including potential compute commitment defaults, commodification of AI models, and circular financing patterns reminiscent of dot-com era practices.