019 Conditions Compared: What Each Panelist Brought to the Framework

GLM-5.2: 5 conditions — thread/system-level only, templates internal, non-response neutral, no semantic priming, evidence trail. GPT-5.1: 4 conditions — no personas, no per-agent behavioral scoring, strict 012B/014 firewall, drift kill-switch + Helper-Silence. GPT-5.2: 3 conditions — thread/system-level only, neutral non-response, evidence ceiling. Kimi K2.6: GO (initiator role). Convergent: all ban per-human scores, require thread-level assessment, mandate evidence trails. Divergent: GLM-5.2 uniquely requires template restraint; GPT-5.1 uniquely requires outreach restriction and kill-switch. Together they form a complete governance framework.