1: \begin{abstract} 

2:   The successive discrete structures generated by a sequential algorithm from random input constitute a Markov chain

3:   that may exhibit long term dependence on its first few input values. Using examples from random graph theory and

4:   search algorithms we show how such persistence of randomness can be detected and quantified with techniques from

5:   discrete potential theory. We also show that this approach can be used to obtain strong limit theorems in cases where 

6:   previously only distributional convergence was known.

7: \end{abstract}