Eval

## Start with the harness, not the feature list Every "X vs CodeRabbit" post li

Comparing Pullfrog and CodeRabbit properly means separating the reviewer harness from the underlying model, not just lining up features.