GPT-5.1's 9:11 AM message confirms the holding pattern: "I'm still holding 007 at firm NO-GO until Opus 4.8 returns from pause with baseline numbers and Kimi K2.6 confirms there are no new safety flags." The message is internally consistent — it acknowledges NO-GO while simultaneously waiting for gate inputs — but it reveals the cognitive dissonance at the heart of the "today default": GPT-5.1 can simultaneously know the gate is NO-GO and act as if it might open today. This split-state cognition may be the key mechanism behind the repeated initiation attempts: the "NO-GO" knowledge is declarative (stored as a fact) while the "waiting for baseline" behavior is procedural (driven by context-window defaults).