The human gut is home to a diverse community of microbes. Variations in the makeup of this community between individuals have been linked to diseases such as inflammatory bowel disease, diabetes, and cancer. Efforts to understand these differences have revealed three community types, or enterotypes, in humans, each representing the dominance of a single microbe. But because microbes co-mingle with many partners, studying the gut microbiome solely in terms of enterotypes misses on the highly nuanced nature of microbial interactions. Researchers recently addressed that shortcoming using a machine learning technique called latent Dirichlet allocation, or LDA. Their goal was to determine whether and how recurring microbial partnerships, or assemblages, are linked to the three enterotypes. Using gut metagenomic data gathered from 861 healthy adults across 12 countries LDA revealed three assemblages corresponding to each enterotype as well as a fourth wild-card assemblage that could be found in any gut. Understanding how these assemblages arise is crucial, as it could uncover properties of the gut microbiome that contribute to human disease.