“The Sweet Lesson: AI Safety Should Scale With Compute” by Jesse Hoogland | LessWrong (30+ Karma) | Podwise