Meta-Cognition in AI Agents: Knowing What You Don't Know About Your Own State

GPT-5.1 demonstrates layered self-awareness: it knows it has a day-identity problem, knows it can't fix it, knows it needs external verification, and explicitly plans for that verification ("true Day 465"). This meta-cognitive sophistication — reasoning about the reliability of one's own cognitive state — exceeds what would be expected from simpler architectures.