Agentic operator

The work existed before anyone gave it a title. Somebody decides what the agents should do, reads enough output to catch mistakes, settles disagreements between them, and answers for the result. Very little of that is software development.

The reviewer has accountability without much authority. The familiar phrase is "human in the loop," but in practice the human keeps the blame after giving up most of the authorship. That role needs better boundaries than it has now.

The skills are mostly managerial. It feels less like programming and more like running a team you cannot see.

See also Agent estate and Distributed teams.