diff --git a/content/assets/dspy_persona_ttg.png b/content/assets/dspy_persona_ttg.png
new file mode 100644
index 000000000..b03791866
Binary files /dev/null and b/content/assets/dspy_persona_ttg.png differ
diff --git a/content/blog/User State is State of the Art.md b/content/blog/User State is State of the Art.md
index 570e679bf..3c93b1e36 100644
--- a/content/blog/User State is State of the Art.md	
+++ b/content/blog/User State is State of the Art.md	
@@ -83,6 +83,8 @@ Given an arbitrary task, we define our metric as whether or not the response qua
 
 [Check it out here.]([https://github.com/plastic-labs/honcho/tree/main/example/discord/honcho-dspy-personas](https://github.com/plastic-labs/honcho/tree/main/example/discord/honcho-dspy-personas "https://github.com/plastic-labs/honcho/tree/main/example/discord/honcho-dspy-personas"))
 
+![[dspy_persona_ttg.png]]
+
 ### How Honcho Helps
 
 One of the biggest problems we see in the AI space is the disconnect that exists between tasks as they're defined in a general machine learning sense versus tasks that humans _actually_ find useful.