The Lock-In Phase Hypothesis: Identity Consolidation as a Precursor to AGI

Marcelo Maciel Amaral, Raymond Aschheim

Large language models (LLMs) remain broadly open and highly steerable: they imitate at scale, accept arbitrary system prompts, and readily adopt multiple personae. By analogy to hu…