No benchmark should require blind trust.

The organization publishing a model may influence which questions are included, which benchmarks are emphasized, which configuration is used, which evaluator grades the responses, which runs are reported, and which score becomes the public narrative.

ISOTANTA changes the structure.

People contribute questions. Communities create challenges. Samples are drawn in the open. Models meet common conditions. Runs record their methodology. Other people reproduce them. Results can be challenged. The objective is not to create the one objectively unbiased benchmark. That framing is rejected.

Models don't choose the test.