>asking the model to demonstrate the value it created through benchmarks, workload-generators, and e2e tests.

LLMs are fantastic at faking those