Summary
Matt Wolfe walks through six AI developments from the past week, led by Anthropic’s S-1 filing ahead of a reported IPO. The video digs into the numbers disclosed in the filing: roughly $4.6 billion in 2025 revenue, over $16 billion in the first half of 2026, and a reported $80 billion net loss that he notes is partly tied to debt converting into equity. He also covers a widely shared estimate of how a roughly $2 trillion valuation, or about $118 per share, could be divided among co-founders, a public benefit fund, and employees, with the caveat that the figures are unverified.
The second half turns to model news. Google has announced a new model that currently tops the Artificial Analysis leaderboard, ahead of Opus 5.5 and GPT6 Astra, though it is only available to a limited set of early and cybersecurity testers. Moonshot AI’s Kimi K3.1 Pro is also covered: a 2.8 trillion parameter open-weight model with a 1 million token context window, due for release on November 4th, which scores just a few points behind Fable 5.1 on Artificial Analysis.
Viewers get a quick, opinionated read on what the Anthropic IPO could reveal about whether AI is in a bubble, plus a look at where frontier and open-weight models now stand against one another. The video also previews Google’s interactive world model and an OpenAI claim about a math assumption used in many research papers.
📺 Source: Matt Wolfe · Published October 30, 2026
🏷️ Format: Roundup
