Associated press release will probably give an overview/context better than me (obviously taken with a grain of salt because it's a press release): https://www.prnewswire.com/news-releases/basecamp-research-launches-basefold-a-breakthrough-in-3d-protein-structure-prediction-of-large-complex-protein-structures-302085262.html
Basically, Basecamp has gone out and partnered with different organisations around the world to make a dataset that they argue covers a wider range of biodiversity than the paltry model-organism-specific sets we normally have. They use this dataset to supplement the MSA during AlphaFold2 inference, and say they get better performance and higher-quality structures out of it.