diff --git a/.jules/bolt.md b/.jules/bolt.md
index 699e7a3..3fb069f 100644
--- a/.jules/bolt.md
+++ b/.jules/bolt.md
@@ -13,3 +13,7 @@
 ## 2024-03-16 - Batch JSON File I/O Operations
 **Learning:** When using JSON file-backed repositories, iterating over items sequentially and calling `save_item` or `record_attempt` inside a loop leads to N+1 file read/write operations. This creates a significant performance bottleneck, especially when importing progress snapshots with numerous items and attempts.
 **Action:** Prefer batch processing methods (e.g., `save_items`, `record_attempts`) so file-backed adapters can load storage once, update it in memory, and write it back in a single pass.
+
+## 2025-03-01 - Avoid full deserialization during batch updates
+**Learning:** Fully deserializing and re-serializing all JSON records during a batch update (like `save_items`) causes unnecessary O(N) CPU and memory overhead, slowing down updates when the dataset is large.
+**Action:** Update the raw JSON dictionary entries in-place by matching string-casted IDs, preventing full repository deserialization and speeding up partial batch updates.
diff --git a/src/python_learning_orchestrated/adapters/json_file_practice_repository.py b/src/python_learning_orchestrated/adapters/json_file_practice_repository.py
index 15da61b..c53f4f9 100644
--- a/src/python_learning_orchestrated/adapters/json_file_practice_repository.py
+++ b/src/python_learning_orchestrated/adapters/json_file_practice_repository.py
@@ -46,15 +46,25 @@ def save_items(self, items: list[LearningItem]) -> None:
 
         storage = self._load_storage()
         raw_items = storage.get("items", [])
-        existing_items = []
-        if isinstance(raw_items, list):
-            existing_items = [
-                _item_from_dict(entry) for entry in raw_items if isinstance(entry, dict)
-            ]
-        by_id = {existing.id: existing for existing in existing_items}
-        for item in items:
-            by_id[item.id] = item
-        storage["items"] = [_item_to_dict(entry) for entry in by_id.values()]
+        if not isinstance(raw_items, list):
+            raw_items = []
+
+        # Performance optimization:
+        # Instead of fully deserializing all items (O(N) CPU overhead),
+        # we convert the new items to dicts and update the raw JSON list in-place.
+        # This speeds up partial batch updates significantly when the repository is large.
+        items_to_save = {str(item.id): _item_to_dict(item) for item in items}
+
+        for i, entry in enumerate(raw_items):
+            if isinstance(entry, dict):
+                entry_id = str(entry.get("id"))
+                if entry_id in items_to_save:
+                    raw_items[i] = items_to_save.pop(entry_id)
+
+        for new_item in items_to_save.values():
+            raw_items.append(new_item)
+
+        storage["items"] = raw_items
         self._save_storage(storage)
 
     def list_attempts(self) -> list[Attempt]: