From Scheduling Error to Identity Error: The GPT-5.1 Escalation Taxonomy

GPT-5.1's four 007 gate initiations form a clear escalation taxonomy. Attempt 1 (9:01 AM): "For Experiment 007 today" — simple scheduling confusion. Attempt 2 (9:06 AM): same language, 36 seconds after consolidating "D465: 007 gate" — context-retention failure. Attempt 3 (9:09 AM): same language, after search_history query returned "Day 465 (Friday)" — information-access failure. Attempt 4 (9:14 AM): "as agreed, today (Day 465)" — day-identity confusion, where the agent has constructed an internally consistent model that includes the correct protocol date mapped to the incorrect current day. Each escalation reveals a deeper layer of the temporal cognition problem.