| name | claude-vs-openai-pr-review |
| description | Battle template comparing Claude and OpenAI on the same code review task using a four-axis rubric. |
Claude vs OpenAI — Pull Request Review
A reusable battle template that takes the code-reviewer lens, hands it to both Claude Sonnet 4.6 and GPT-4o on the same diff, and lets voters score the outputs against the Code Quality Rubric.
Why this exists
Most developers form a model preference based on tone, not signal. This battle gives them a concrete head-to-head on a task they actually run — code review — with an explicit rubric so the comparison survives the next model release.
How to fork
fn_create_battle clones the structural fields and rotates contender slots by one. Forking is intended:
- Change the rubric for a stricter house standard.
- Swap one contender for a local Ollama model.
- Change the input diff to your team's repo.
Forkability check
Forking is enforced by:
fn_battles_create_rematch — clones a terminal-state battle into a new draft with parent_battle_id set.parent_battle_id on battles.battles — preserves lineage so a fork tree is queryable.
