I'll discuss a
number of insights into the process of nonlinear structure formation which come
from the study of random walks crossing a suitably chosen barrier. These
derive from a number of new results about walks with correlated steps, and
include a unified framework for the peaks and excursion set frameworks for
estimating halo abundances, evolution and clustering, as well as nonlinear,
nonlocal and stochastic halo bias, all of which matter for the next generation
of large scale structure datasets.