I don’t understand this video. Stroustrup compares two algorithms with O(n) operations (linked lists, where search is linear, vs arrays, where insertion/deletion is linear) for repeatedly inserting/deleting in a sorted list, poses a silly question (“For what n is a linked list better?” …which makes no sense because there is no reason to think one algorithm would ever “overtake” the other), and gets the obvious answer (the “constant” factors for arrays are much better, hence they win for all n).
To (eventually) beat arrays, which always have great constant factors (because of their spatial locality, simplicity, and lack of pointer-chasing), you need to have a data structure that has better asymptotic complexity for your problem. For the problem he poses, of maintaining a sorted list, this would be something like a red-black tree.
For ordinary linked lists (single or double) to beat arrays, you need to have a problem that requires repeated insertion/deletion in the middle of the list, but which requires no linear-time searching (i.e. where you have a pointer to the necessary list element already for some reason).