diff --git a/README.md b/README.md
index 79939ab..78fc2bd 100644
--- a/README.md
+++ b/README.md
@@ -79,23 +79,6 @@ Arms:
 
 Voluntary adoption is itself a metric: if grep fails to find the deleted file, does the agent reach for findtest? That directly tests the mechanism.
 
-```mermaid
-flowchart TD
-    COMMIT["real maintainer commit\npost-cutoff · 3 codebases"] --> WK["isolated git worktree\n.git link stripped"]
-    WK --> DEL["delete associated test file\nv2 deletion protocol"]
-    DEL --> SPLIT((" "))
-    SPLIT --> A1["A1 — control\nRead · Grep · Glob · Bash"]
-    SPLIT --> A2["A2 — treatment\nRead · Grep · Glob · Bash\n+ findtest MCP (voluntary)"]
-    A1 --> GEN1["generated test\n# target file: declared"]
-    A2 --> GEN2["generated test\n# target file: declared"]
-    GEN1 --> JUDGE["LLM judge\nblinded pairwise"]
-    GEN2 --> JUDGE
-    GEN1 --> METRICS["AST alignment · location · taste"]
-    GEN2 --> METRICS
-    JUDGE --> RESULT["win-rate · Δalignment\nadoption rate · per codebase"]
-    METRICS --> RESULT
-```
-
 ### Three codebases
 
 The study ran across three repos chosen to span a complexity gradient: