How do you know whether your prompt is bad or the model sucks? How do you know whether the model sucks because it doesn't have enough parameters or because it's badly trained or has a bad lora?

All without trying all possible combinations of models - at least yourself.