TL;DR audience_programming.mp4 rather than using the most frequent string in chat, this algorithm finds a substring prefix which maximizes the following formula: $LN(N-2)$ where L is the length of the substring and N is the number of occurence of substring