Deploy analyst-buddy (Gradio app + serving)

server/app_ui.py

@@ -155,13 +155,14 @@ _BUILD_PROOF = (
     "(pet-shop retail: 0% → 42%)."
 )
 _BUILD_WIP = (
-    "This work is ongoing and still improving: one- and two-table questions are "
-    "reliable, three-table joins are still a weak spot, and a longer training run "
-    "showed signs of overfitting that we're still analyzing."
+    "Current status: one- and two-table questions are reliable, three-table joins "
+    "are still a weak spot, and a longer training run showed signs of overfitting "
+    "that we're analyzing."
 )
 _HEAD2HEAD_INTRO = (
-    "Don't take the number on faith — watch the same tiny model, same question, "
-    "before and after fine-tuning."
+    "A side-by-side on the same question: the same tiny model before and after "
+    "fine-tuning. Off-the-shelf, it gives up or guesses wrong. Fine-tuned, it "
+    "explores the data and answers."
 )
 # The "what we're building" body (rendered inside the always-DARK card by
 # ``_why_card_html`` below, optionally with the head-to-head showcase nested in).