Researchers reported new experiment-guided approaches to overcome limitations in AI-based protein and genome design workflows, including improved AlphaFold guidance with experimental data and new methods for precise integration of large DNA sequences. In Nature Biotechnology, teams described ways to guide AlphaFold models using experimental constraints to better capture conformational heterogeneity rather than collapsing structures into a single predicted shape. On the genome engineering side, additional Nature Biotechnology studies describe retargeted serine integrases and recombinases engineered for one-step, precise insertion of large DNA payloads in human cells. These advances focus on retargeting established integration enzymes to reduce reliance on multi-step processes while maintaining locus specificity. The direction is notable for translational gene therapy development, where efficient and precise delivery of larger genetic constructs can determine whether a platform can scale beyond small edits. While the reports emphasize method performance and constraints consistency, their ultimate value will depend on robustness in relevant cell types and delivery contexts. Together, the work expands the toolchain linking experimental measurements with AI-guided models and provides more precise options for large-DNA integration strategies.