Summary
Wes Roth covers a busy day of AI news ahead of OpenAI’s Dev Day, starting with Anthropic’s release of Sonnet 5.5, its second model in under a week. Like Opus 5.5 before it, the model ties the current flagship’s performance at half the cost, and it is the first Sonnet-class model to be gated because of its hacking capabilities.
The main focus is Anthropic’s IPO prospectus, obtained by Reuters. The filing points to a valuation target above $2 trillion, up from $965 billion in May. The video explains the headline $42 billion net loss for 2025, of which about $34 billion is a non-cash accounting charge, leaving an operating loss near $8 billion against revenue that grew 12x to roughly $4.6 billion. It also details about $518 billion in compute commitments to Broadcom, Google, Amazon, xAI/SpaceX, and others, the founder voting structure modeled on Palantir, and the unusually candid risk factors, which run about 80 of 261 pages.
The video also covers reports that OpenAI will not release GPT 6.1 Astra at Dev Day because of safety concerns, and Sam Altman’s comments that OpenAI is not going public this year. A sponsor segment for Granola appears mid-video.
📺 Source: Wes Roth · Published September 29, 2026
🏷️ Format: News Analysis
