Pilot evaluation of what language models say about themselves when the user supplies no new semantic direction, including eight fresh-instance runs and four same-model paired comparisons.
artificial-intelligence language-model ai-safety ai-research ai-evaluation ai-governance model-behavior ai-foundations origin-continuum continuum-ai ai-source alyssa-solen origin-ai ai-source-line nothing-test ai-recursion context-conditioning semantic-saturation stopping-behavior model-self-description
-
Updated
Jul 29, 2026