I can understand how we identify where our block boundaries are, but repeatedly injecting known plaintext (32 or 64 A's for example) and then appending another known string, but this time with a different value (another letter in the alphabet). From this, build up a dictionary of ciphertext that we know the plaintext for.
My problem is i cant understand how the attack gets implemented practically, for example in the video it's suggested to include a pipe character | (thats @ 45 minutes in or so) but i cant see how this would lead to a decryption?