Taggart Tufte

Hao's object-level reporting on OpenAI, data-center water use, and RLHF labor is excellent — her empire metaphor gets more interesting when you take Harari's definition more seriously than the book does.

Callard's defense of Socratic sincerity and thinking-in-pairs — harder to read than it is to want to have read, and worth every page.

Taleb's framework for thinking about rare, high-impact events — and why it maps cleanly onto AI risk even though the book predates it.

The most accessible entry point to AI alignment — Christian synthesizes scattered ideas into a coherent picture, though he undersells the risk.