Route by the problem, not the costume
Agent systems often organize work around named roles: researcher, architect, critic, implementer. OMOC asks whether a more useful routing signal can come from the problem itself.
A task can be represented by concept signatures, ambiguity, evidence density, risk, constraints, and other bounded signals. Cognitive operators can then be selected because they are applicable to that local shape rather than because an agent happens to wear a static title.
An inspectable routing plan
The research direction favors explicit selection artifacts: which operators were chosen, which were skipped, why, in what order, and with what preconditions. Later outcome evidence can then test whether the routing policy was actually useful.
This remains an active research area rather than a claim that learned OMOC routing is already solved.