docs: qualify README claims for intellectual honesty

- Change "Core Innovation" to "Core Approach" (more accurate)
- Change "key differentiator" to "explores" (less marketing)
- Correct accuracy figure (46.7% -> 100%, not 33% -> 100%)
- Add context that all 45 tasks share same navigation entry point
- Link to publication roadmap for methodology and limitations
- Change "No technical expertise needed" to "Reduced prompt engineering"

The goal is accuracy over marketing appeal.

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

Author: abrichr

README.md

@@ -285,18 +285,18 @@ flowchart TB
     class L0,L1,L2 implemented
 ```
 
-### Core Innovation: Demo-Conditioned Prompting
+### Core Approach: Demo-Conditioned Prompting
 
-OpenAdapt's key differentiator is **demonstration-conditioned automation** - "show, don't tell":
+OpenAdapt explores **demonstration-conditioned automation** - "show, don't tell":
 
 | Traditional Agent | OpenAdapt Agent |
 |-------------------|-----------------|
 | User writes prompts | User records demonstration |
 | Ambiguous instructions | Grounded in actual UI |
-| Requires prompt engineering | No technical expertise needed |
+| Requires prompt engineering | Reduced prompt engineering |
 | Context-free | Context from similar demos |
 
-**Retrieval powers BOTH training AND evaluation**: Similar demonstrations are retrieved as context for the VLM, improving accuracy from 33% to 100% on first-action benchmarks.
+**Retrieval powers BOTH training AND evaluation**: Similar demonstrations are retrieved as context for the VLM. In early experiments on a controlled macOS benchmark, this improved first-action accuracy from 46.7% to 100% - though all 45 tasks in that benchmark share the same navigation entry point. See the [publication roadmap](docs/publication-roadmap.md) for methodology and limitations.
 
 ### Key Concepts