> complex things are just that: complex, and modeling them is at least as complex, so it will always be hard, I recall reading that a measure of the complexity

> complex things are just that: complex, and modeling them is at least as complex, so it will always be hard, I recall reading that a measure of the complexity of a thing (formally) is the shortest possible non-lossy description. This suggests that your insight is correct. The question is how much further can we go in automating parts of the descriptions.