Docs updates to clarify optimize usage.

Author: qhoxie

docs/src/content/docs/optimization.mdx

@@ -74,8 +74,19 @@ Optimization can significantly improve accuracy on real-world tasks:
 
 ## Persisting Results
 
-Saving/loading an optimized extractor is not yet implemented.
-For now, re-run `optimize()` when you start up, or persist your training data and configuration.
+Save and load optimized extractors to reuse them without re-running optimization:
+
+```python
+# Save after optimization
+extractor.save("./my_extractor")
+
+# Load later
+from langstruct import LangStruct
+loaded = LangStruct.load("./my_extractor")
+
+# Use immediately - optimization is preserved
+result = loaded.extract("new text")
+```
 
 ## Advanced (If You Need It)
 
@@ -110,25 +121,25 @@ extractor.optimize(
 ## Common Questions
 
 **Q: Do I always need training data?**
-A: No! Optimization can work without training data, but providing examples improves results significantly.
+A: You need example texts, but not necessarily expected outputs. If you don't provide `expected_results`, LangStruct uses the LLM's confidence ratings to optimize. Providing expected outputs significantly improves accuracy.
 
 **Q: How long does optimization take?**
 A: Usually 1-5 minutes for typical datasets (10-100 examples).
 
 **Q: Can I optimize an already optimized extractor?**
-A: Yes! You can keep optimizing with new data as you get it.
+A: Yes, you can continue optimizing with new data as you collect it.
 
 **Q: Will this make my extractions slower?**
-A: No - optimization happens once during training. Production extraction speed is the same.
+A: No - optimization happens once during training. Production extraction speed is unchanged.
 
 **Q: What happens when I switch models?**
-A: Just change the model and re-optimize! Same training data, same accuracy - zero prompt rewriting needed.
+A: Change the model and re-optimize with the same training data. No prompt rewriting needed.
 
 ## Next Steps
 
 <CardGrid>
 	<Card title="Try It Now" icon="laptop">
-		Create a LangStruct extractor and enable optimization when you need accuracy!
+		Create a LangStruct extractor and enable optimization when you need accuracy.
 	</Card>
 	<Card title="Source Grounding" icon="document">
 		[Track where information comes from](/source-grounding/)

docs/src/content/docs/quickstart.mdx

@@ -87,9 +87,8 @@ extractor = LangStruct(example=schema)
 # See optimization in action
 extractor.optimize(
     texts=["training texts..."],
-    expected=[{"expected outputs..."}]
+    expected_results=[{"expected outputs..."}]  # Optional - uses confidence if omitted
 )
-print(f"Optimized accuracy: {extractor.score:.1%}")
 ```
 
 ## Process Multiple Documents (with quotas)

docs/src/content/docs/why-dspy.mdx

@@ -119,7 +119,7 @@ extractor = LangStruct(example={
 
 # 2. Let MIPROv2 optimize prompts and examples automatically
 extractor.optimize(
-    training_texts=["Apple reported $125B in Q3...", "Meta earned $40B..."],
+    texts=["Apple reported $125B in Q3...", "Meta earned $40B..."],
     expected_results=[
         {"company": "Apple", "revenue": 125.0, "quarter": "Q3"},
         {"company": "Meta", "revenue": 40.0, "quarter": "Q3"}
@@ -148,15 +148,15 @@ extractor = LangStruct(
     example={"company": "Apple", "revenue": 100.0},
     model="gpt-5-mini",
 )
-extractor.optimize(training_texts, expected_results)
+extractor.optimize(texts=training_texts, expected_results=expected_results)
 
 # 6 months later, switch to Claude - just two lines!
 extractor.model = "claude-3-7-sonnet-latest"
-extractor.optimize(training_texts, expected_results)  # Auto-reoptimizes prompts
+extractor.optimize(texts=training_texts, expected_results=expected_results)  # Auto-reoptimizes prompts
 
 # Or use local models for privacy
 extractor.model = "ollama/llama3.2"
-extractor.optimize(training_texts, expected_results)  # Works the same way
+extractor.optimize(texts=training_texts, expected_results=expected_results)  # Works the same way
 
 # Same accuracy, zero prompt rewriting, zero vendor lock-in
 ```