Decentralized architectures are increasingly adopted to improve resilience and data sovereignty by shifting trust away from centralized authorities to auditable software. This research targets coordination-free decentralized systems based on local-first principles, where users store data locally, selectively share it, and retain full control over access and updates. Conflict-free Replicated Data Types (CRDTs) enable such systems by resolving conflicts without requiring coordination, but existing work largely assumes cloud environments with fixed, trusted replicas. We aim to make CRDTs practical in decentralized, coordination-free settings characterized by unstable membership, non-uniform trust, and resource-constrained nodes by pursuing four objectives. First, we plan to design novel causality-tracking mechanisms tailored to highly dynamic topologies. Second, we will rethink CRDT abstractions to support partial replication. Third, we intend to develop a resilient framework for privacy and integrity. Finally, we aim to investigate memory optimizations for resource-constrained devices.