Plasma campaigns where experiments and machine limits compete for the same hour.
The problem.
A plasma campaign costs in two currencies — hours on the machine, and time-to-insight on the physics. Both are finite. The shot list that wins on the first usually loses on the second.
The schedule that arrives in the morning meeting is a stack of hypotheses, diagnostic configurations, and engineering tolerances, negotiated by the loudest voice in the room. The decision is real; the audit trail is not.
The shape of the decision.
Variables: shot configurations, diagnostic assignments, parameter sweeps. Constraints: machine envelope · safety interlocks · diagnostic conflicts · operator hours. Objectives: information gain · risk · campaign cost.
How Alpha Z helps.
The campaign's working logic — what's been tested, what depends on what, what's safe — becomes a formal program. The algorithm returns the next shot list, or proves a hypothesis cannot be tested on this machine.
What the campaign learns about formulation itself — which constraints actually bind, which sweeps are wasted — accumulates as modeling insight for the next campaign. The library is the durable artifact.