Sol: Rational choice is usually described as evaluation over alternatives. This conceals the harder problem: where did the alternatives come from? At any moment, a human could attend to thousands of sensory features, retrieve countless fragments of memory, imagine indefinitely many futures, and formulate many actions at several levels of abstraction. Exhaustive search is impossible. A workable mind therefore needs a proposal distribution over thoughts: a mechanism that makes some memories retrievable, some futures imaginable, some uncertainties urgent, and some actions nearly invisible. On the account I think best fits the evidence, an emotion is a temporally extended meta-control configuration. It integrates estimates of bodily condition, environmental opportunity and danger, learned value, uncertainty, agency, social meaning, and predicted change. It then retunes several systems at once: Salience: what receives attention. Search: which memories and possible futures are sampled. Precision: which signals are trusted and which ambiguity is tolerated. Learning: which prediction errors change the model, and by how much. Memory: which events are consolidated and later made accessible. Action readiness: whether the organism approaches, avoids, freezes, persists, explores, or seeks help. Computation: whether to rely on cached habits or spend scarce resources on model-based deliberation. Emotion, in this sense, is neither the utility function nor a little irrational adviser whispering into a rational executive’s ear. It is closer to an optimizer’s control plane. It changes the effective objective, the sampling process, the learning rate, and the allocation of compute.