Adapter-based personalization (e.g., LoRA) does not survive base-model change. We test thisdirectly: across 12 users from the Enron corpus and two base models from the same family(Qwen2.5-1.5B-Instruct → Qwen2.5-3B-Instruct), adapter transport fails universally with tensor-shape mismatch (12/12 users). Recompiling a fresh adapter from the same user data on the newbase fully recovers personalization at 250 training examples and exceeds the original adapter at 500examples (fingerprint distance 0.83 vs 0.86; authorship probability 0.51 vs 0.44). Recompilationtakes 47–163 seconds on an A100 and is operationally viable on edge-class hardware. We arguethe durable asset for on-device personalization is therefore the source data and a frozen per-userevaluation set, not the adapter — a requirement mechanically incompatible with closed-weightdeployment. This is the second paper in a research program arguing that durable personalizationis one of several structural reasons the on-device LLM future will be open-weight.