Over the past year, we've made the transition from a traditional API/SPA split to a much more technically challenging server-rendered SPA with full cache handover. In this talk, we'll present pitfalls, what we learned and how we optimized compute spend to make sense for our workload.