Most existing continuous optimization methods take "candidate solution points" as the basic operation objects, generating the movement trajectory of solutions in the search space through gradient descent, random perturbation, or group interaction mechanisms. However, this point-centered search paradigm generally suffers from premature convergence, local optimum locking, and high sensitivity to step size and hyperparameters in highly non-convex, strongly constrained, or structurally complex problems. This paper proposes a novel continuous optimization method—the Boa Optimization Algorithm. This method no longer regards the optimization process as the movement of points in space, but redefines it as a process of active geometric reconstruction and topological collapse of a feasible region during the search. The algorithm uses a dynamic feasible body as the core state and performs gradual half-space pruning of the search space through a directional survival pressure assessment mechanism, thereby achieving geometric contraction of the optimal solution region without relying on individual optimal updates or explicitly setting exploration and utilization phases. This paper systematically elaborates on the method from the aspects of problem modeling, algorithm mechanism, mathematical formalization, and theoretical properties, proving that it has important properties such as monotonic volume contraction, directional consistent convergence, and robustness to noisy objective functions. This method provides a unified geometric-physical perspective for continuous optimization, which differs from traditional point search.