Summary Rapid convergence that can be prescribed by a user is appealing for many applications with high requirements. This stimulates finite‐time stabilization with arbitrarily prescribed settling‐time and so‐called prescribed‐time stabilization. But their continuous realizations had to restrict uncertainties and/or to bear truncated run of controllers. This paper, for nonlinear systems with large uncertainties, realizes not only the convergence within the prescribed time, but also the non‐truncated run. First, by lending finite‐time stabilization to prescribed‐time stabilization and integrating dynamic compensation, an adaptive controller with time‐varying components is devised such that the system state reaches the origin at a finite time less than the prescribed time, while exhibiting local asymptotic stability (of the origin). Then by monitoring the finite time online, the time‐varying components of the adaptive controller are frozen as their values at the finite time. The asymptotic stability guarantees the frozen adaptive controller can make the system state remain at the origin for all future time. But the above finite time could not be detected in practice, due to ubiquitous disturbances. We thus modified the detection to ensure that the system state enters a vicinity of the origin before the prescribed time and stays there afterwards under some conditions on uncertainties and disturbances. Two simulation examples illustrate the effectiveness of the proposed controller.