Written from the same material as the essays, with room for the arguments and the cases that do not fit in one.
An operation is made of people, workflows, decisions, data and technology, and what it does is decided by how those are wired together rather than by the quality of any one of them. This is the long form of that argument: how to see the arrangement rather than the pieces, and what changes once you can.
A consultant is accountable for a recommendation. A product engineer is accountable for shipped capability. Neither owns whether the operation changes. This is the discipline that does: the method, the unit of delivery, the exit condition, and the cases that do not fit inside an essay.
Most agentic systems fail on operating conditions rather than on model capability. This is about the conditions: where autonomy carries real leverage, where it multiplies an existing problem, and what has to be true of the surrounding system before any of it holds in production.