diff --git a/services/openclaw/deployment.yaml b/services/openclaw/deployment.yaml
index 19ab81aa..d54fd9dc 100644
--- a/services/openclaw/deployment.yaml
+++ b/services/openclaw/deployment.yaml
@@ -20,7 +20,7 @@ spec:
         app: openclaw
       annotations:
         ai.bstein.dev/role: testing-triage
-        ai.bstein.dev/placement: Jetson pool (titan-20/21)
+        ai.bstein.dev/placement: Jetson MVP lane (titan-21)
     spec:
       serviceAccountName: openclaw-triage
       automountServiceAccountToken: true
@@ -36,7 +36,6 @@ spec:
                   - key: kubernetes.io/hostname
                     operator: In
                     values:
-                      - titan-20
                       - titan-21
       initContainers:
         - name: init-config
diff --git a/services/openclaw/ollama-deployment.yaml b/services/openclaw/ollama-deployment.yaml
index 49c0736f..08498ba1 100644
--- a/services/openclaw/ollama-deployment.yaml
+++ b/services/openclaw/ollama-deployment.yaml
@@ -20,7 +20,7 @@ spec:
         app: openclaw-ollama
       annotations:
         ai.bstein.dev/model: qwen2.5:1.5b-instruct-q4_0
-        ai.bstein.dev/gpu: Jetson pool (titan-20/21)
+        ai.bstein.dev/gpu: Jetson MVP lane (titan-21)
     spec:
       runtimeClassName: nvidia
       affinity:
@@ -31,7 +31,6 @@ spec:
                   - key: kubernetes.io/hostname
                     operator: In
                     values:
-                      - titan-20
                       - titan-21
       volumes:
         - name: models
@@ -69,8 +68,8 @@ spec:
               memory: 4Gi
               nvidia.com/gpu.shared: 1
             limits:
-              cpu: "4"
-              memory: 10Gi
+              cpu: "3"
+              memory: 8Gi
               nvidia.com/gpu.shared: 1
       containers:
         - name: ollama
@@ -102,10 +101,10 @@ spec:
             timeoutSeconds: 5
           resources:
             requests:
-              cpu: "2"
+              cpu: "1"
               memory: 8Gi
               nvidia.com/gpu.shared: 1
             limits:
-              cpu: "6"
-              memory: 12Gi
+              cpu: "3"
+              memory: 10Gi
               nvidia.com/gpu.shared: 1