I think it makes sense, but wanted to make sure that the .cdf_ tails the tail into consideration, whereas .ldf_ does not?

This is the intended behavior. On the cdf_, all labels are Age-Ult, e.g. 12-Ult, 24-Ult, etc. to denote that they are the cumulative of the available ldf_.

At this point in the example, no tail estimator has been fitted (and they are optional). Without a tail estimator, it is assumed that there is no tail (or more technically a tail = 1.0). If you do fit a tail, it will show up in the cdf_ and ldf_ vectors in the ways you'd expect.

Fun fact - you can to dev.ldf_.incr_to_cum() to calculate the cdf_ as well. The incr_to_cum method recognizes that the ldf_ triangle are a set of patterns.

Ahh yes, I think I read it wrong the first time.

And what a neat feature with dev.ldf_.incr_to_cum(), I'll add that to the tutorial.