As recently recognised by the Nobel Prize committee, machine-learning enabled protein structure prediction is a transformative technology that has the potential to revolutionise molecular biology. However, big-tech chose not to systematically apply these approaches to viral proteins, consequently the impact of tools like AlphaFold in virology has yet to be fully realised. Motivated by an interest in the structures and functions of viral fusion glycoproteins, we have been applying protein structure prediction first to dozens, then hundreds and now thousands of viral species. In doing so, we have gained insights on fusion mechanisms, and revealed new perspectives on the complex and interweaved evolutionary history of viruses. I will give an overview of what we have achieved so far and consider where the field may go in the future.