Paper Proposes Runtime-Independent Persistent Agents

AI researcher Elvis Saravia highlighted the paper Runtime-Independent Persistent Agents. Its authors argue that agents are typically described by the model and harness running at one moment. That framing works for a single session yet leaves long-lived agents underspecified when they switch models or servers over months. The work presents a runtime-independent architecture meant to keep identity, memory, and code intact regardless of the underlying model, harness, or host.