This field study grounds proactive coding-assistant design in real developer behavior rather than lab-only task completion.
The agent proactively surfaces code-quality suggestions, and the study analyzes when developers accept, ignore, or experience those suggestions as disruption.
The paper observes developers over five days in an IDE setting and analyzes interaction timing, workflow phase, acceptance, and perceived value.
The deployment period is still short, and the study focuses on coding rather than broader computer-use workflows.
Use this for evidence about workflow boundaries, timing calibration, and real-world acceptance of proactive developer tools.