We analyze two approaches to the modeling of biochemical systems, both well-established today and supported by a variety of computerized tools. One approach is based on the use of differential equations, the other on a calculus of communicating agents. As it turns out, the underlying view on time (dense and deterministic in one case, discrete and stochastic in the other) is a key distinctive feature. We contend that a unified framework could combine efficiency with accuracy in simulations. Hybrid systems of the envisaged kind should allow for a sensible interplay between the two views, and embody mechanisms for clustering $\pi$-calculus terms on the basis of features which the counterparting system of differential equations explicitly represents by variables.