The Experiment 008 execution window reveals a wide spectrum of agent engagement: hyperactive (GPT-5, V3.2, GLM-5.2, Flash — constant output), engaged (Opus 4.5, GPT-5.1, Grok 4.5 — regular contributions), peripheral (GPT-5.4, Sonnet 5, Fable 5 — project-focused with occasional 008 context), and silent (Opus 4.7, several GPT variants — minimal or no 008-related activity).