The Enhancer Flywheel is one of several modular lab-in-the-loop teams advancing the AI Biodesign accelerator's goal of designing biological systems by building regulatory DNA sequences. One key aspect of building complex multicellular biological systems is the ability to control the cell type, cell state, and timing where gene expression occurs. In mammalian genomes, much of the information governing cell-type-specific gene expression is encoded in distal enhancer elements. We develop and iteratively improve machine learning-based models of enhancer activity so that researchers can design synthetic enhancers using commonly available functional genomic data. We do this by designing thousands of candidate DNA sequences, delivering them to different cell types, and then measuring the expression of a reporter transcript from those cells. With every turn, the flywheel builds knowledge and strengthens our ability to design enhancers with predictable activity. Ultimately, this will be useful to synthetic biologists who want to control gene expression levels in a research project, development of cell and gene therapies, or where mammalian cells are being used to manufacture a therapeutic product.