[tjpick]

Original source FOR the "orders of magnitude" claim seem to usually point back to Fred Brooks Mythical Man Month (5-10x) or Robert Glass (Facts and Fallacies of Software Engineering) 28x. I don't think either of these multipliers are accurately expressed as "orders of magnitude", regardless of them being supported by evidence.

There is some discussion of the topic in "Making Software: What Really Works, and Why We Believe It."

[tellarin]

Hmmm. I agree that 'orders' is not an appropriate choice of words.

But I usually don't interpret this kind of claims literally, so I took the GP remark as saying that in practice (IRL) there was no such 'wide gap in productivity', which is opposite to claims/data I saw before (as you mention, 5-20x) and to my own experience.