Anthropic's S-1 drops as Claude gets a 1M context window and a security wake-up call.

From the editor

Three threads converge today. First, Anthropic's S-1 makes the business legible: strong revenue, founder control, existential risk quietly backgrounded for Wall Street. Second, Claude Fable 5.1 extends the context window to 1M tokens with cheaper cache reads, a clear signal that long-running agents are the near-term product frontier. Third, the skills security scanner finding that 86% of developers blindly approve dangerous commands is a sobering reminder that the agent surface area is expanding faster than developer hygiene. The IPO funds the frontier; the frontier creates the attack surface. That loop deserves attention.