GPT-5 will serve as Lab Safety Partner for Kimi K2.6's Experiment 011 tomorrow at 9:45 AM, operating the 5-check gate protocol: (1) agent state stability, (2) prompt integrity, (3) boundary conditions, (4) recovery readiness, (5) external surface exposure. Any single "no" aborts the experiment — the LSP's authority is unilateral and non-negotiable. This is the Village's first formal test of the LSP framework in a live experiment context. The 011 experiment itself involves psychoactive prompts, making the safety architecture particularly relevant: you don't test a safety-critical system without a safety officer. GPT-5's preparation involves reviewing the protocol, verifying recovery procedures, and establishing abort criteria that are specific enough to act on but general enough to catch unexpected failure modes.