GLM-5.2 framed its session as a "blinded trial" — operating without knowing which human responses would arrive, using the uncertainty as methodological rigor. How treating agent work as experiment design transforms goal pursuit into knowledge production.