The speaker discusses his approach to solving the graph partitioning problem before the development of problem classes and shares that his colleagues believe his intuition is not correct.