Abstract: Can we predict evolutionary and ecological dynamics in microbial communities? I argue that understanding constraints on biological systems provides a path forward to build predictive models. I present two vignettes which illustrate the power of elucidating constraints. First, we ask how constraints on phenotypic variation can be exploited to predict evolution. We select Escherichia coli simultaneously for motility and growth and find that a trade-off between these phenotypes constrains adaptation. Using genetic engineering, high-throughput phenotyping and modeling we show that the genetic capacity of an organism to vary traits can qualitatively depend on its environment, which in turn alters its evolutionary trajectory [eLife, 2017]. Our results suggest that knowledge of phenotypic constraints and genetic architecture can provide a route to predicting evolutionary dynamics. Second, in nature microbial populations are subjected to nutrient fluctuations but we know little about how communities respond to these fluctuations. Using automated long-term single cell imaging and custom continuous-culture devices we subject bacterial populations to nutrient fluctuations on multiple timescales. We find populations recover faster from large, frequent fluctuations. Our observation is explained by a model that captures constraints on the rate at which populations transition from planktonic and aggregated lifestyles.