diff --git a/nodes.md b/nodes.md
new file mode 100644
index 0000000..25655a0
--- /dev/null
+++ b/nodes.md
@@ -0,0 +1,37 @@
+# ⚠️ Running List of (Potentially) Problematic Nodes
+
+
+|      Node       |         Reference          |
+| :-------------: | :------------------------: |
+| `x4002c2s1b0n0` | 🚨[^slow-vaino-2025-08-23] |
+| `x4717c7s2b0n0` | 🚨[^slow-vaino-2025-08-23] |
+| `x4514c6s1b0n0` | 🚧[^slow-vaino-2025-08-23] |
+| `x4208c0s5b0n0` | 🚧[^slow-vaino-2025-08-23] |
+| `x4311c0s5b0n0` | 🚧[^slow-vaino-2025-08-23] |
+| `x4102c4s5b0n0` | 🚧[^slow-vaino-2025-08-23] |
+| `x4314c7s0b0n0` | 🚧[^slow-vaino-2025-08-23] |
+| `x4608c3s1b0n0` | 🚧[^slow-vaino-2025-08-23] |
+| `x4711c3s6b0n0` | 🚧[^slow-vaino-2025-08-23] |
+
+[^slow-vaino-2025-08-23]:
+    
+    <details closed><summary>[<b>2025-08-23</b>] (Väinö Hatanpää, on Slack):</summary>
+
+    - [Report](https://cels-anl.slack.com/archives/C058HKVJ0QL/p1755957183678639):
+      
+      > I did some full machine PyTorch benchmarking and there were some
+      > troublemaker nodes.  
+      > These were >10% slower than others (sample size 1-3, could be randomness,
+      > but some were reoccurring):
+      >
+      > ```bash
+      > x4514c6s1b0n0 x4208c0s5b0n0 x4311c0s5b0n0 x4102c4s5b0n0 x4314c7s0b0n0 x4608c3s1b0n0 x4711c3s6b0n0
+      > ```
+      >
+      > And these two were hanging with some communication calls, every time I tried:
+      >
+      > ```bash
+      > x4002c2s1b0n0 x4717c7s2b0n0
+      > ```
+
+    </details>