When building an Android app locally, many Gradle tasks are read from the cache after the first build. In contrast, when building the app on a remote CI/CD machine, no cached task outputs exist because the machine starts from a clean state every time. In those cases, cached task outputs can be read from a remote server by using fingerprints based on task inputs. Remote caching has been shown to reduce CI build times for most projects. However, the benefits vary from project to project and using the remote cache in a meaningful way for a large project can be a trial and error exercise. This talk aims to reveal how remote caching works by starting from a simple Gradle project and adding one layer of complexity at a time until reaching a minimal Android app. It is meant to demystify a couple of typical experiences with the remote build cache for Gradle: "It doesn't work and I don't know why." and "It works and I don't know why."