tamte
because stamp() function is only valid during the cooking of the for subnet or old copy sop for that matter, you can't see the evaluated value as it's different every iteration so when observed it's not cooking at that moment and it falls back to the value specified as third argument, so in your case 0
new for loops allow you to observe a single iteration, by default is the last one, but you can pick a specific one so thats why detail() expression is able to resolve the actual value when observed as the metadata node still holds the values that belong to chosen iteration
otherwise your result in viewport seems to be the same, if thats the case, it appears to be working correctly during cooking
(also 3rd argument in detail() is component index, while in stamp() it's the mentioned fallback value, so while they may seem similar when both written with 0, they mean very different things)
Thanks for the quick reply. Yes this makes sense. I was actually able to get all the way through the first volume of this tutorial, but now I'm having an issue with only certain seeds working, and I've been comparing the two networks wondering if this could be the cause. But, since the viewport output is the same, this seems more unlikely.
As I move through both networks, the other thing that seems like it might be an issue is in the closed balcony section with the treat_body loop. I wonder if I replaced the opinputpath expression correctly by referring directly to the node which the previous opinputpath index 1 seemed to refer (find_closed_balc). They evaluate the same, both as 0, but in this case it seems to make all the geometry disappear.
Here's an example of my network and evaluation:


Here's an example of Anastasia's network and evaluation:
