The discourse on AI and jobs has been almost entirely wrong because it assumes a uniform operator. There is no uniform operator. There are operators who can write a clear brief and operators who cannot, and the gap between them is widening by the quarter.
The operators who can write a brief — define the constraints, the success criteria, the acceptable failure modes — get 5x output from the same models. The operators who cannot get a slightly faster version of the same mediocre work they were already shipping. The model is the same in both cases.
The new core skill
Sequencing, not prompting. Knowing which decisions can be delegated to a model, which need a human in the loop, and which need a meeting. That is a judgment skill. It compounds. It is also almost entirely absent from how most companies are hiring for AI roles right now.
"AI doesn't replace operators. It removes the cover that bad operators used to hide behind."
The unkind truth is that the operators who feel most threatened by AI right now are usually the ones whose value-add was the friction the tooling just removed. That is not a bug in the tools. It is a signal.
Dispatch Newsletter
One operator-grade breakdown every Tuesday. SEA tech, AI economics, execution frameworks. No fluff.
