If you read their description again, [I can't be positive, but] I think they are not using the word "dynamic" in the sense of AutoCAD's term "dynamic Block," but in the sense of visibly seeing the result dynamically [in this case, as they drag the cursor around for a rotation], on top of building the rotation into something so they don't need to go through all the steps to copy and rotate manually.  Making multiple copies in the same drawing and being able to rotate each separately as it's copied does not sound to me like something one would ever want to do with North Arrows.

 

But a confirmation from the OP of whether I am correctly interpreting their intent would be helpful.

 

*^C^CCOPY \ INS \\ROTATE L ;@ \

 

If the Blocks you would do this with do not contain anything in them that has its own insertion point [such as Text or nested Blocks], you can have it automatically get the insertion point for you, and save you a step:

 

*^C^CCOPY \ INS @ \ROTATE L ;@ \

 

But if they do contain anything nested that has an insertion point, when you pick the Block on that nested part, it will find the nested part's insertion point, rather than the overall Block's insertion point [or even if you pick it on some other part but the nested part is within Object-Snap range and its insertion point is closer than the overall Block's].  So either use the first macro with its extra step, or be careful about where you pick the Block.

 

Also note that the rotation of the copy will start the same as that of the one selected to Copy, and the Rotate operation will go from there, not from the Block's own zero rotation.  So you'd do best by having one at zero rotation that you can select repeatedly to make multiple copies, rather than [for example] each time selecting the last copy you made.

 

They both depend on the User successfully selecting something, and on its being a Block [actually, it would work with other object types, but might have unexpected results for things that don't have insertion points], and on its not being on a locked Layer.  AutoLISP would be able to make allowance for the possibility that you might miss, or pick the wrong kind of object, or pick something on a locked Layer.