This is a Plain English Papers summary of a research paper called Work Agents: A Framework for Persistent, Open-Ended Tasks and Verifiable Outcomes. If you like these kinds of analyses, you can find more AI and machine-learning research on AIModels.fyi or follow us on Twitter.
A work-centric definition of digital agents
Work persistence and work openness are the paper’s two dimensions for describing the conditions under which a work agent operates. The framework treats an agent as a system that combines a foundation model with a runtime harness, acts in a stateful workspace, and is judged by whether it delivers a verifiable outcome rather than merely generating useful text.
The distinction shifts attention from model capability alone to the full execution system. A foundation model may propose actions, while the harness supports execution across tools and applications; the workspace retains state that can change as work proceeds. The abstract does not specify a particular model, harness implementation, or benchmark used to instantiate this system view. The framework treats an agent as a system that combines a foundation model with a runtime harness, acts in a stateful workspace, and is judged by whether it delivers a verifiable outcome rather than merely generating useful text.
