diff --git a/.jules/bolt.md b/.jules/bolt.md
index ba84fca0..bf784525 100644
--- a/.jules/bolt.md
+++ b/.jules/bolt.md
@@ -21,3 +21,7 @@
 ## 2025-05-30 - O(1) Blockchain Hash Lookup
 **Learning:** Calculating a blockchain-style hash chain requires finding the previous record's hash. A naive DB scan is O(log N) with indexes, but high-concurrency follow operations can be optimized.
 **Action:** Implement an in-memory thread-safe cache for the last hash of each grievance to achieve O(1) lookup during follower creation, falling back to DB on cache miss.
+
+## 2025-07-15 - Fast Bounding Box Pre-filter
+**Learning:** Calculating great circle distance (Haversine) for every issue against a target location is computationally expensive (O(N) with heavy math ops like sin, cos, atan2). In high-traffic aggregations, this can become a bottleneck.
+**Action:** Use a fast bounding box pre-filter (`get_bounding_box` with a 5% epsilon) to quickly discard issues that are definitely outside the search radius before running the expensive exact haversine distance calculation.
diff --git a/backend/spatial_utils.py b/backend/spatial_utils.py
index 8af329a3..213bfeaa 100644
--- a/backend/spatial_utils.py
+++ b/backend/spatial_utils.py
@@ -76,10 +76,24 @@ def find_nearby_issues(
     """
     nearby_issues = []
 
+    # Fast bounding box pre-filter with a 5% epsilon to optimize expensive haversine calculations
+    epsilon_radius = radius_meters * 1.05
+    min_lat, max_lat, min_lon, max_lon = get_bounding_box(
+        target_lat, target_lon, epsilon_radius
+    )
+
     for issue in issues:
         if issue.latitude is None or issue.longitude is None:
             continue
 
+        # Skip issues outside the bounding box
+        lon_delta = (issue.longitude - target_lon + 180) % 360 - 180
+        lon_half_width = max(target_lon - min_lon, max_lon - target_lon)
+        if not (min_lat <= issue.latitude <= max_lat and (
+            abs(target_lat) >= 89.9 or abs(lon_delta) <= lon_half_width
+        )):
+            continue
+
         distance = haversine_distance(
             target_lat, target_lon,
             issue.latitude, issue.longitude