This landing lab runs a simplified numerical rocket model. It is not validated flight software. The saved nominal state starts at three thousand two hundred meters, descending at one hundred sixty five meters per second. The controller estimates braking distance, decides when to ignite, and applies bounded throttle and steering commands. We will follow those exact source rules rather than its claims about real rocket architectures. At the initial mass of forty three thousand eight hundred kilograms, the source takes ninety five percent of maximum thrust per unit mass, then subtracts gravity. That gives about eight point five two meters per second squared of available deceleration. It divides speed squared by twice that value: about one thousand five hundred ninety eight meters. Adding a fixed twenty five meter buffer gives an initial ignition threshold near sixteen hundred twenty three meters. The threshold is recomputed as the state evolves. The squared speed term is the crucial sensitivity. Holding the assumed mass and deceleration fixed, doubling the descent speed multiplies braking distance by four. Here the shorter bar represents sixteen hundred meters approximately, and the longer bar sixty four hundred. This is a comparison of the formula, not a prediction of another flight. Actual simulation steps also change velocity, mass, drag and orientation. Once the engine is lit, the throttle command is clipped between forty and one hundred percent. In this illustrative clamp comparison, a requested zero point two becomes zero point four, while a requested one point two becomes one point zero. Horizontal errors form a desired pitch clipped at one quarter radian, and gimbal commands are clipped near five degrees. Bounds limit commands; they do not guarantee a safe landing. Velocity is advanced before position with a zero point zero two second numerical step. Start, pause, change wind or sea heave, and export the recorded samples to inspect the model. Each animation frame advances one numerical step while running, so a labeled fifty hertz step is not proof of actual wall clock execution at fifty hertz. The CPU timing display is generated from a sine expression, not a measured benchmark. The sensor noise checkbox is stored but does not enter these physics equations. Read those interface labels separately from demonstrated computation.