Researchers announced two advances that move AI‑driven genomics toward practical design: BGI’s Genos, a human‑centric genomic foundation model trained on hundreds of near telomere‑to‑telomere human assemblies, and a separate team that used large genomic language models to design functional genes across prokaryotes. Genos aims to predict functional elements and variant pathogenicity with human‑focused training data; the semantic design work showed models can generate novel gene sequences with measurable activity. Both efforts emphasize training on high‑quality, diverse human assemblies or massive prokaryotic datasets to capture long‑range genomic context and functional motifs. The papers suggest foundation models can accelerate variant interpretation, functional element discovery, and the de novo design of genetic parts for therapeutics and synthetic biology. Clarification: a genomic foundation model is a large AI trained on raw DNA sequences to learn patterns that can be repurposed for downstream genomics tasks such as variant effect prediction or sequence design.