This talk will be a mixture of some past research and some future research that I have done in the name of capturing long duration speaker specific characteristics. First, I will speak of my attempts to parameterize Dan's sequential non-parametric system. The aspiration of the research was to develop a framework that enabled the system to learn the optimal recipe for combining scores. For those who don't remember Dan's system, there will be a brief overview. Afterwards, I will talk about a new system, currently under development, which hopes to combine the power and simplicity of the GMM system with the untapped potential of time-varying speaker idiosyncrasies.