This talk examines how coding agents can move beyond individual assistance and become trustworthy contributors to production software workflows. It shares lessons from CI-based experiments, including separating generation from verification, setting safety boundaries, and deciding when an autonomous system should stop and return control to a human.