In order to greatly increase the power density of boiling liquid reactors, more turbulent and effusive boiling of the moderator coolant must ensue. However, this would entail handling very large random reactivity excursions with its attendant dangers. Perhaps, this problem could be circumvented by a novel, hyper-speed control comprised of “rod equivalent” systems of very fast response. This would allow the reactor to approach its stability limit more closely and thereby increase the power density. To realize such systems, this effort is directed toward a different conceptualization of the reactor control problem as opposed to the less than adequate small excursion linearized theory extant. The idea involved in “bang-bang” control is that of ever driving the reactor toward its equilibrium state as rapidly as possible from randomly perturbed states in which it finds itself because of the turbulent moderator. The control problem is formulated in a fashion analogous to the brachistochrone class of problems, but with a stochastic feature due to the random reactivity fluctuations. Using the methods of dynamic programming, a functional equation in the minimum time for the reactor to be driven back to equilibrium is obtained. From this is derived an optimal reactor control policy. A controller computer can then be synthesized which instantaneously senses the perturbed state of the reactor. It then computes the optimal reactivity policy and sends actuating signals to the “rod(s)” system. The responding reactor is then found in its new perturbed state, which is again read, etc. This procedure continually drives the reactor toward the equilibrium state in the sense of minimum time defined above.