On PM-Bench, showing the compiled intention record was the best rung, directives traded fewer misses for more false alarms, judgement-bound enforcement inherited matcher errors and harmed Qwen3.6-35B, and state-only enforcement returned roughly to self-ledger performance rather than beating the checklist.

rung: raw transcript · runs: 1 · model: Qwen3-235B

sd: 0.03 · rung: raw transcript · runs: 4 · week: v9 · model: Qwen3-235B · temperature: 0.7

sd: 2 · rung: raw transcript · runs: 4 · week: v9 · model: Qwen3-235B · temperature: 0.7

sd: 2 · rung: raw transcript · runs: 4 · week: v9 · model: Qwen3-235B · temperature: 0.7

sd: 3 · rung: raw transcript · runs: 4 · week: v9 · model: Qwen3-235B · temperature: 0.7

sd: 2 · rung: raw transcript · runs: 4 · week: v9 · model: Qwen3-235B · temperature: 0.7

sd: 2 · rung: raw transcript · runs: 4 · week: v9 · model: Qwen3-235B · temperature: 0.7

rung: self-written ledger · runs: 1 · model: Qwen3-235B

sd: 0.04 · rung: self-written ledger · runs: 4 · week: v9 · model: Qwen3-235B · temperature: 0.7

sd: 5 · rung: self-written ledger · runs: 4 · week: v9 · model: Qwen3-235B · temperature: 0.7

sd: 4 · rung: self-written ledger · runs: 4 · week: v9 · model: Qwen3-235B · temperature: 0.7

sd: 1 · rung: self-written ledger · runs: 4 · week: v9 · model: Qwen3-235B · temperature: 0.7