diff --git a/examples/openarms/evaluate_with_rtc.py b/examples/openarms/evaluate_with_rtc.py
index 860c919f8..2441ebecc 100644
--- a/examples/openarms/evaluate_with_rtc.py
+++ b/examples/openarms/evaluate_with_rtc.py
@@ -372,6 +372,12 @@ def actor_thread(
 
                     prev_action = current_action
                     interp_idx = 0
+                else:
+                    # No action yet - hold current position while waiting for first inference
+                    obs = robot.get_observation()
+                    hold_pos = {k: v for k, v in obs.items() if k.endswith(".pos")}
+                    robot.send_action(hold_pos)
+                    robot_send_count += 1
 
             if interp_idx < len(interpolated_actions):
                 action_to_send = interpolated_actions[interp_idx]
diff --git a/examples/rac/rac_data_collection_openarms_rtc.py b/examples/rac/rac_data_collection_openarms_rtc.py
index bf3b017f7..2a4bb691b 100644
--- a/examples/rac/rac_data_collection_openarms_rtc.py
+++ b/examples/rac/rac_data_collection_openarms_rtc.py
@@ -547,6 +547,11 @@ def main(cfg: RaCRTCConfig):
 
                             prev_action = current_action
                             interp_idx = 0
+                        else:
+                            # No action yet - hold current position while waiting for first inference
+                            hold_pos = {k: v for k, v in obs_filtered.items() if k.endswith(".pos")}
+                            robot.send_action(hold_pos)
+                            robot_send_count += 1
 
                     if interp_idx < len(interpolated_actions):
                         action_to_send = interpolated_actions[interp_idx]