fix(backend): reduce ultrathink value from 65536 to 60000 for Opus 4.5 compatibility (#1173)

The hardcoded ultrathink value of 65536 exceeded Claude Opus 4.5's
max_output_tokens limit of 64000, causing all Ultra+Opus tasks to fail
with API Error 400.

Changes:
- apps/backend/phase_config.py: Reduced ultrathink from 65536 to 60000
- apps/frontend/src/shared/constants/models.ts: Mirrored backend change
- tests/test_thinking_level_validation.py: Updated test assertions

The new value of 60000 provides a 4k buffer under Opus 4.5's limit,
allowing the SDK to add its overhead token without exceeding the max.

Refs: ACS-295

Co-authored-by: StillKnotKnown <stillknotknown@users.noreply.github.com>

apps/backend/phase_config.py

@@ -25,7 +25,7 @@ THINKING_BUDGET_MAP: dict[str, int | None] = {
     "low": 1024,
     "medium": 4096,  # Moderate analysis
     "high": 16384,  # Deep thinking for QA review
-    "ultrathink": 65536,  # Maximum reasoning depth
+    "ultrathink": 60000,  # Maximum reasoning depth (must be < Opus 4.5's 64000 limit)
 }
 
 # Spec runner phase-specific thinking levels

apps/frontend/src/shared/constants/models.ts

@@ -28,7 +28,7 @@ export const THINKING_BUDGET_MAP: Record<string, number | null> = {
   low: 1024,
   medium: 4096,
   high: 16384,
-  ultrathink: 65536
+  ultrathink: 60000 // Maximum reasoning depth (must be < Opus 4.5's 64000 limit)
 } as const;
 
 // ============================================

tests/test_thinking_level_validation.py

@@ -35,7 +35,7 @@ class TestThinkingLevelValidation:
 
     def test_ultrathink_max_budget(self):
         """Test that 'ultrathink' returns maximum budget."""
-        assert get_thinking_budget("ultrathink") == 65536
+        assert get_thinking_budget("ultrathink") == 60000
 
     def test_invalid_level_logs_warning(self, caplog):
         """Test that invalid thinking level logs a warning."""
@@ -91,4 +91,4 @@ class TestThinkingLevelValidation:
         assert get_thinking_budget("low") == 1024
         assert get_thinking_budget("medium") == 4096
         assert get_thinking_budget("high") == 16384
-        assert get_thinking_budget("ultrathink") == 65536
+        assert get_thinking_budget("ultrathink") == 60000