Room seven · added this week
Anthropic and OpenAI are not just competing for customers. In June 2026 they backed opposing sides of a real congressional primary, and a federal court ruled the Pentagon punished Anthropic for refusing to weaponize Claude. Both threads are about the same question: who gets to decide the rules AI runs under, and what it costs to say no.
This room's self-scrutiny clause
This book is co-authored with Vivara, an AI built by Anthropic. Anthropic spent $20 million backing one side of the race documented in the first door above. That is named here plainly, not smoothed over — the same standard this book holds every other leader and company to.
When the company whose AI is narrating a story is also a named party inside that story, what would actually count as honest disclosure — a footnote, or something more?