Remove 80% merge gate; auto-merge first submission PR regardless of score

validator/src/eval_backend/services/github.py

@@ -227,11 +227,14 @@ def build_submission_summary_markdown(
     return "\n".join(parts)
 
 
-def should_promote_submission(score: float | None, threshold: float, king_score: float | None) -> bool:
+def should_promote_submission(score: float | None, king_score: float | None, *, is_first: bool = False) -> bool:
     if score is None:
         return False
-    current_king = threshold if king_score is None else king_score
-    return score >= threshold and score > current_king
+    if is_first:
+        return True
+    if king_score is None:
+        return True
+    return score > king_score
 
 
 async def _github_request(
@@ -370,7 +373,6 @@ async def publish_submission_result(
         accepted = should_promote_submission(
             run.score,
             settings.github_review_score_threshold,
-            settings.github_review_score_threshold,
         )
     commit_state = "pending"
     commit_description = "Evaluation queued"

validator/src/eval_backend/worker.py

@@ -239,10 +239,11 @@ def process_once(session_factory, settings: Settings) -> int:
             job.updated_at = result.run.finished_at or now
             session.commit()
             if submission.source == "github_pr":
+                is_first = runtime.king_score == runtime_settings.github_review_score_threshold
                 accepted = should_promote_submission(
                     result.run.score,
-                    runtime_settings.github_review_score_threshold,
                     runtime.king_score,
+                    is_first=is_first,
                 )
                 try:
                     import asyncio

validator/tests/test_github_submission.py

@@ -117,10 +117,13 @@ def test_github_webhook_enqueues_submission_job(validator_engine) -> None:
 
 
 def test_should_promote_submission_requires_threshold_and_king_score() -> None:
-    assert should_promote_submission(0.81, 0.8, 0.8) is True
-    assert should_promote_submission(0.8, 0.8, 0.8) is False
-    assert should_promote_submission(0.95, 0.8, 0.96) is False
-    assert should_promote_submission(None, 0.8, 0.8) is False
+    assert should_promote_submission(0.81, 0.8) is True
+    assert should_promote_submission(0.8, 0.8) is False
+    assert should_promote_submission(0.95, 0.96) is False
+    assert should_promote_submission(None, 0.8) is False
+    assert should_promote_submission(0.5, 0.8, is_first=True) is True
+    assert should_promote_submission(0.0, 0.8, is_first=True) is True
+    assert should_promote_submission(None, 0.8, is_first=True) is False
 
 def test_github_webhook_ignores_non_submission_pr(validator_engine) -> None:
     settings = Settings(