In section “Neural network implementation in Python”, in the 1st video “Forward prop in a single layer” at 4:27 there is a typo (repeated in the Practice quiz):
w2_1 = np.array( [-7, 8] ) must be a 1D vector with 3 elements (not 2 as shown).
The 2 element vector can not legally dot-product with a 1D vector with 3 elements:
a1 = np.array( [a1_1, a1_2, a1_3] )