In the past 48 hours, the AI community has witnessed a nationwide self-testing frenzy triggered by a mysterious prompt, as OpenAI and Anthropic have almost simultaneously fallen into an "intelligence degradation scandal." OpenAI was exposed for secretly conducting a gray-scale test of GPT-5.6 on the Codex platform and secretly reducing users' thinking budget. Opus 4.8 has suffered an epic downgrade, with the once-amazing model now frequently failing at the most basic logical reasoning and even starting to PUA users; Opus 4.8 Max has been severely criticized by users as "having its brain cut off," with its performance plummeting from amazing to rock-bottom, even worse than the old Haiku model. The AI community discovered that OpenAI might be conducting a small-scale gray-scale test of GPT-5.6-sol, where some conversations that were supposed to run GPT-5.5 xhigh were secretly routed to an unknown model named "gpt-5.6-sol." Actual test data shows that a normal, full-powered gpt-5.5 xhigh should return a Juice result of 768 when faced with specific test instructions, but for users routed to the gpt-5.6-sol gray-scale test pool, the returned value dropped precipitously to 128. The value has shrunk by a factor of 6 from 768 to 128, which could mean either that the inference efficiency of GPT-5.6 has achieved