A general question about a model's behaviour by uniformly selecting a finite number of inputs from an infinite input space.

Question

I am conducting an assessment of a model to draw conclusions about its behaviour. I uniformly select 100 natural numbers as inputs. By feeding these 100 inputs to the model, I obtain 100 corresponding outputs, based on which I draw conclusions. One could argue that my conclusions are limited to these 100 inputs, and using different inputs from the infinite space of natural numbers might lead to different conclusions. However, in response, I argue that the uniform sampling approach ensures that regardless of the specific inputs chosen, the conclusions drawn regarding the model's performance remain consistent. Is my argument correct? If yes, I am seeking a few academic references on this topic.