The Explosion Afraid of Itself
A geometric braking mechanism on recursive self-modification for Active Inference agents, arguing that radical and runaway self-improvement meets intrinsic curvature limits.
Raman, not Varelse — Why Acceptably Controllable AGI Policies Cluster Around Human-Like Stacks
The AGI architectures that humans will tend to find affordably verifiable and robustly assurable will be those that are epistemically close enough to themselves.